Manipulation of self-produced footstep sounds:

Influences on the perception of one’s own body

weight, motor behavior and emotions

Maria Basia

Project report submitted in part fulfilment of the requirements for the degree
of Master of Science (Human-Computer Interaction with Ergonomics) in the
Faculty of Brain Sciences, University College London, 2014

NOTE BY THE UNIVERSITY

This project report is submitted as an examination paper. No responsibility can be
held by London University for the accuracy or completeness of the material
therein.

1
 ACKNOWLEDGMENTS

I am sincerely and heartily grateful to my Supervisor Ana Tajadura-Jimenez.

Without her patience, professionalism, valuable guidance and contribution this
study would not have been possible. I would also like to thank my co-supervisor
Nicolai Marquardt for his valuable help and advice. Additionally, I would like to
thank Dr Nadia Berthouze for her insightful view and advice and Andrew
Symonds for providing us with the EMG equipment.

I would like to express my gratitude to people in the Institute of Philosophy for

kindly allowing me to carry out the experimental study in their lab and especially
Merle Fairhurst and Ophelia Deroy for their help and advice. Also I would like
to thank all the participants that took part in my study.

A great thanks goes to my fellow students for being great and supportive
throughout this intensive academic year. I owe earnest thankfulness to my family
and especially my parents for always supporting and encouraging me. Without
their support I would not have been able to attend this MSc Programme.
Thanks to Stelios for always being there for me.

2
 ABSTRACT

Body movements generate sounds which are significant to our everyday

interactions. It comes with no doubt that walking plays a fundamental role in our
everyday lives and footstep sounds constitute an invaluable source of
information about people’s activities, perceived body characteristics and
immediate environment. The effects of the manipulation of specific auditory
cues in the perceived sound source characteristics have been demonstrated in
past studies. Moreover, emerging theories have indicated that walking related
sounds can affect people’s auditory perceived characteristics of a walker’s body,
motor behavior and emotions. However, the effects of the real-time,
manipulated, self-produced footsteps sound on the perception of one’s own body
and the related motor behavior and emotions have not yet been explored.

This aim of this study is to investigate whether the provision of real-time,

manipulated, self-produced footsteps sounds can alter people’s perception of
their own body weight, motor behavior and emotions during walking. For this
purpose, a system comprised of a pair of sensored shoes prototype and audio
equipment was built to provide participants with auditory feedback and measure
their behavioral changes.

In a controlled experiment participants were asked to walk down an 8.5 m

corridor while wearing the prototype and were being provided with real-time
footsteps sound through headphones. The sounds furnished were self-produced
and they were either participants’ amplified natural footsteps sounds or their
footstep sounds manipulated in terms of their frequency spectrum. A number of
different sensors were used to capture participants’ behavioral and physiological
changes. A web-based application was utilized and questionnaires were also
administered to collect data regarding participants perceived body weight, speed,
strength and posture.

The study’s suggests that people felt lighter when the high frequency
components of their footsteps sound were selectively increased (ranges between
1-4 kHz) and heavier when the footsteps sound’s low frequency components
were selectively amplified (ranges between 63-250 Hz). Their motor behavior

3
(e.g. acceleration when lifting foot of the ground) and effective responses
(valence, arousal, and dominance) were also altered. A series of design
recommendations for the design of this kind of applications are also provided.

The findings along with the extensive use of wearable devices and the
integration of multiple sensors which record body activity into smartphones
suggest excellent possibilities for the use of manipulated footsteps sound in the
design of technology for fitness and rehabilitation, virtual reality and games.

Word count: 14.929

4
Contents
1 INTRODUCTION ....................................................................................................... 7
2 LITERATURE REVIEW, RESEARCH QUESTION AND HYPOTHESES.............................. 8
2.1 Literature Review ............................................................................................. 8
Human Gait .............................................................................................................. 8
Sound as a Source of Information: The Perceptual Attributes of Sound.................. 9
Sound and Perception of Body .............................................................................. 12
Sound and Motor Behavior.................................................................................... 14
Sound and Emotions .............................................................................................. 16
Implications for HCI ............................................................................................... 17
2.2 Research Question and Hypotheses .............................................................. 19
3 PROTOTYPE AND METHODS .................................................................................. 23
3.1 Prototype ....................................................................................................... 23
Prototype and System Description ........................................................................ 23
Sensor Setup .......................................................................................................... 26
Exploration ............................................................................................................ 29
3.2 Methods ........................................................................................................ 31
Participants ............................................................................................................ 31
Materials and Apparatus ....................................................................................... 31
Design .................................................................................................................... 39
Procedure .............................................................................................................. 40
3.3 Data Extraction .............................................................................................. 42
Behavioral and Physiological Data ......................................................................... 42
4 DATA ANALYSIS...................................................................................................... 48
4.1 Questionnaires............................................................................................... 50
Perceived body weight as measured by the 3D Body Visualization Application .... 50
Task Experience Questions .................................................................................... 51
Self-Assessment Manikin (SAM) ............................................................................ 58
Spanner Questionnaire .......................................................................................... 61
Self-Efficacy Questionnaire .................................................................................... 62
4.2 Behavioral and Physiological Measures ......................................................... 64
Pressure sensors .................................................................................................... 64
Accelerometer ....................................................................................................... 68
Galvanic Skin Response – GSR ............................................................................... 70

5
5 DISCUSSION, FUTURE WORK AND LIMITATIONS ................................................... 74
5.1 Results Discussion .......................................................................................... 74
5.2 Applications in HCI and Design recommendations ........................................ 83
Applications in HCI ................................................................................................. 83
Design recommendations ...................................................................................... 83
5.3 Limitations and Future Research ................................................................... 87
6 Conclusion ............................................................................................................. 89
REFERENCES .................................................................................................................. 90
APPENDICES................................................................................................................... 95

6
 1 INTRODUCTION

Previous researches have showed that the input of multi-sensory modalities is

essential for the performance of even the simplest everyday activities. Body
movements generate sounds which are fundamental to our everyday interactions.
Over the past years there has been a significant research effort to investigate the
perceptional attributes of sound (e.g Lederman,1979; Leed,1993). Moreover,
manipulations of specific sound cues, such as the frequency spectrum, have been
proved to alter the perceived characteristics of the sound source (e.g.Jousmäki
and Hari,1998; Guest et al.,2002). Considering the fundamental role of walking
in our everyday interaction, researchers have also provided evidence that pre-
recorded walking related sounds can affect the perceived characteristics of a
walker’s body such as weight (Li et al.,1991) and posture (Pastore et al.,2008)
while other theorists (Tajadura-Jimenez et al.,2012) suggested that sound can
alter the mental representation of our body. It has also been demonstrated that
walking related sounds alter people’s motor behavior (Bresin et al.,2010) and
evoke affective responses (Tonetto et al.,2014). While past studies provide
indication that walking related sounds can alter perceived body characteristics,
motor behavior and emotions, no study so far has explored the effects of real-
time, self-produced, manipulated footstep sounds on the perception of one’s own
body weight and the related motor behavior and emotions. It is the aim of this
study to further explore the specific domain. The study also aims to provide
design recommendations for the use of footsteps sounds in the design of audio-
based technologies. For the purpose of the research a system consisting of audio
equipment and sensors was build.

In this thesis Chapter 2 reviews the literature on the basics of gait mechanism,
the perceptual attributes of sound and sound effects on body perception, motor
behaviour and emotions. Implications for HCI are also presented. The research
question and hypotheses are also presented in the light of the literature review
findings. Chapter 3 presents the prototype and the methodology applied in the
experiment. Chapter 4 presents the study results. Chapter 5 discusses the results,
presents the design recommendations, implications for HCI, limitations and
future work. Finally, chapter 6 concludes the study.
7
 2 LITERATURE REVIEW, RESEARCH QUESTION AND
HYPOTHESES

2.1 Literature Review

Walking and sound are fundamental to our everyday lives. The present study
aims to investigate the implications of footstep auditory cues in the perception of
our own body weight, walking behavior and emotions. Consequently, in this
chapter we review the literature regarding the basics of gait mechanism, the
nature of the auditory cues elicited as a result of the contact of foot with the
ground and the effects of acoustic stimuli on areas such as perception, gait
behavior and human affective states.

Human Gait

Human gait is a periodic movement of each of the lower limbs from one position
of support to the next (Vaughan et al., 1992; Perry, 1992). The time interval
between two successive contacts of the same foot with the ground defines the
gait cycle (Cunado et al., 2003). The gait cycle is divided into two phases, the
Stance and the Swing phase. According to Vaughan et al., (1992) the Stance
phase begins with the “heel strike” of one of the feet and is followed by “foot
flat”, during which the plantar surface of the foot is in contact with the ground,
“midstance” when the contralateral limb passes the leg that is on the ground,
“heel-off” when heel leaves the ground and “toe-off” which indicates the end of
Stance phase and during which the foot loses contact with the ground. The
Swing phase is comprised by the following events. The “acceleration” when the
flexor muscles are activated to accelerate the foot forward, the “midswing”
which occurs simultaneously with the “midstance” of the contralateral limb and
the “deceleration” when the foot decelerates to prepare for the next heel strike.

The force applied by the lower limb on the ground is a net force F which can be
presented as a spectrum with frequency variations (Visell et al., 2009). While the
low frequency components of this spectrum are related with the Ground
Reaction Force (GRF) and they mostly depend on the walker’s weight and
8
walking rate, the high frequency components are linked to factors such as the
impact between the foot and the ground, the sliding friction and the different
degrees of contact with the ground and they depend on the ground and shoe
materials (Ekimov and Sabatier, 2006). These factors are responsible for the
aural cues elicited during walking.

Walking pattern highly varies across individuals. This is attributed to many

factors (e.g. age, gender, and weight) (Troje, 2008). Past studies have
investigated the association of specific human attributes such as weight and age
with walking patterns and styles. Troje, (2008) in his study on the information
contained in human gait patterns collected motion data from 80 participants
using a motion capture system. Based on these data he developed an application
which displays an animated point-light walker
(http://biomotionlab.ca/Demos/BMLwalker.html). In this application a number
of settings enable the user to change walker’s characteristics such as the body
weight in a light-heavy scale and observe the changes in the gait pattern and
walking posture. For instance, it is observed that lighter bodies appear to have a
higher lifting acceleration of foot and to adopt a straighter walking posture,
whereas heavier bodies show a decreased lifting acceleration of foot and adopt a
stooped/hunched posture.
The present study is concerned with the effects of the acoustic cues emanated
during walking in motor behavior (pressure applied on the floor and acceleration
of feet).

Sound as a Source of Information: The Perceptual Attributes of Sound

Auditory cues play a substantial role in human perception providing invaluable

information about ourselves, the actions we perform and properties of the
external world. Humans have the ability to immediately identify the source of
naturally produced sounds of a wide range (Ballas, 1993). The sounds elicited
during our everyday interactions with the world provide us with valuable
information about the sound source’s characteristics (such as material, solidity
and mass), location (Jenkins, 1985; Norman, 1998; Gaver, 1993) and
environment.

9
The variety of environmental sounds is remarkably wide and it is quite difficult
to generalize the applicability of empirical results in entire sound categories.
Thus, a number of studies have focused on investigating the relationship between
the acoustic properties of specific types of auditory cues and human acoustic
perception (Hardness of percussion mallets (Freed, 1993), classification of
bouncing and breaking bottle events (Warren & Verbrugge, 1984), configuration
of clapping hands (Repp, 1987), surface texture (Lederman, 1979; Lederman et
al., 2002).

Building on past studies on the perceptual properties of sound, researchers have

also investigated the effects of manipulating specific sound properties, such as
the frequency spectrum, on human perception. Jousmäki and Hari (1998)
demonstrated that both the increment of high frequency components (above 2
kHz) and of the overall amplitude of the sound elicited when rubbing two hands
together increased the perceived smoothness/dryness of the skin. However, it
should be noted that the results of the specific study cannot be clearly
comprehended since the assessment scale used combined two measures that
should be assessed independently (roughness and wetness) (Suzuki et al. 2008).
Based on this observation, Guest et al. (2002) replicated the aforementioned
experiment and measured “wetness” and “dryness” separately. Their results
showed that both perceived skin roughness and dryness increased when the
amplitude of the high frequencies was increased. An increment of dryness
perception was also reported when overall sound was amplified.

In the same experiment Guest et al. (2002) also investigated whether

manipulating the frequency spectrum of auditory cues produced by touching
sand paper affects its perceived abrasiveness. The high frequency components
(bands between 2-20 kHz) of the touch-produced sounds were either increased or
decreased. Overall, their results suggest significant effects of frequency
alterations in perceived tactile sensation.

Similarly, Zampini & Spence (2003) showed that the perception of the crispness
and freshness of crisps being eaten was increased when either the overall sound
generated when biting the crisps was amplified or the high frequencies were
increased (ranges from 2-20 kHz).

10
The above mentioned studies indicate the potential effects of manipulating the
frequency components of sound on human perception, providing evidence that
manipulating the frequency spectrum of footsteps sound may lead to analogous
effects.

The Informative Nature of Footstep Sounds

Footstep sounds as body activity related sounds belong to the Self-representation

sound category (Väljamäe et al., 2008). Not only are these auditory cues
fundamental to our navigation and interaction with the environment but also they
are associated with the perception of specific attributes of both the sound source
and the walking event itself. However, only a few studies have further explored
the perceptual characteristics of walking related auditory cues. More specifically,
walking sound source properties that have been investigated in previous studies
include the walker’s gender and weight (Li et al., 1991, Giordano and Bresin,
2006), posture (Pastore et al., 2008), emotions (Giordano and Bresin, 2006), the
type of the ground (Giordano et al., 2012), and the abrasiveness and size of the
walker’s shoe (Giordano and Bresin, 2006).

Li et al. (1991) in their study explored the acoustic properties of walking

generated sounds and their contribution to the way a listener perceives the
gender of a walker. Participants showed high identification ability to judge a
walker’s gender from pre-recorded footstep sounds. Their results demonstrated
that the spectral peak and the high-frequency components are the main
influential sound attributes that affect gender judgment. Their findings also
demonstrate that weight and height are the two main anthropomorphic
characteristics that influence the perceived gender classification and that when
the spectral mode of the footstep sounds was changed to lower frequencies (at
least 125 Hz), the judgment of maleness increased from 50% to 90%.
Accordingly, the perception of femaleness increased to 90% when the spectral
mode was changed to higher frequencies (over 1000 Hz).

Giordano and Bresin (2006) demonstrated that participants showed a high ability
to recognize gender, emotions, weight and shoe properties (shoe size and sole

11
abrasiveness) of a walker. However, a greater accuracy was reported in the
identification of gender, weight and shoe characteristics than in emotions.
Similarly to Li et al. (1991) the results demonstrated that gender classification is
based on spectral peak and high frequency components.

In a more recent research Pastore and colleagues (2008) investigated the ability
of participants to identify the posture (upright-stooped) of a walker based on the
auditory stimuli produced by the walker’s footsteps. While their findings reveal
that both upright and stooped walking elicits complex and variable acoustic
characteristics they do provide an indication that pace and spectral amplitude (in
the range of 100-500 Hz) may be associated with the listeners’ judgments
regarding posture.

Finally, Giordano et al. (2012) in their research investigated the ability of

participants to differentiate ground materials (solid and aggregate) based on
different feedback modalities (auditory, haptic, combination of auditory-haptic
and kinesthetic). The results showed that in all four conditions the two types of
material were rarely confused. However, haptic feedback led to more accurate
assessments than auditory feedback. While this study suggests that auditory
feedback is not effective enough to convey information regarding ground
materials, the study is not statistically significant since only 5 participants were
recruited.

While the above mentioned studies give valuable insights about the
characteristics of the footstep sounds of a walker and the way they are perceived
by humans they do not investigate the provision of real-time self-generated
footstep sounds and their possible affects in perception of one’s one body,
behavior and emotions.

Sound and Perception of Body

Auditory feedback plays a substantial role in the way we perceive our body
(Kitagawa and Spence, 2006). Despite the fact that many study findings
demonstrate that body-representations are continuously updated and affected by

12
sensory inputs such as vision, touch and proprioception (Botvinick and Cohen,
1998; De Vignemont et al. 2005; Haggard et al. 2007) the possible effects of
self-generated sounds in the way humans perceive their bodies have only
recently been studied (Tajadura-Jimenez et al. 2012, Senna et al., 2014).

As mentioned above, Jousmäki and Hari (1998) demonstrated that sound can
alter the perception of skin properties. Building on studies like this one, Senna et
al., (2014) showed that acoustic stimuli (sound of a hammer hitting marble)
provided in synchrony with the slight hitting of participants’ arm with a hammer
altered the auditory perceived material of the arm, resulting in participants
reporting that they feel their arm stiffer and heavier like it was made of marble.

In Tajadura-Jimenez et al’s., (2012) study participants were asked to tap on a

surface with their hand and the perceived location of the sound’s origination,
was manipulated in three different conditions (no alteration, double distance,
four times the distance of tapping location).Their results showed that real-time
manipulation of action-related sounds generated when people were tapping on a
surface, influences the perceived location of the action (tapping). This can be
associated to changes in perceived location of the tapping hand and perceived
arm length.

Building on this, Furfaro et al., (2013) found that accompanying the tapping of a
real or virtual surface with real-time tapping sounds which correspond to the
sound elicited by different degrees of tapping strength, influences perceived
body strength, emotions, tapping behaviour and the perception of the surface
properties (hardness).

In a more recent study, Tonetto et al., (2014) demonstrated that the provision of
different combinations of sounds produced from a variety of soles (leather,
propylene) and ground (carpet, ceramic) materials affect both emotions and
sensations of body, such as feeling “at ease”, “relaxed”, “comfortable”,
“resentful” and “contend”.

While the above described studies suggest that sound affects the perception of
our own body, the effects of self-produced walking generated sounds on
perceived body properties such as weight have not yet been investigated.

13
Sound and Motor Behavior

The effects of sound on human locomotion have been researched and

documented in the recent years with most of the research focusing on musical
stimuli.

Styns et al. (2007) focused their research on the exploration of how music affects
human gait. The participants were provided with either musical or metronome
feedback both ranging on tempo and were asked to adjust their gait tempo to the
stimuli’s tempo. The results revealed that participants adopted a faster walking
pace with the musical stimuli than with the metronome stimuli. This was
attributed to the fact that the information conveyed by music, such as the sound
of specific musical instruments increased the walker’s energy. Additionally,
Moens et al’s., (2010) study provided evidence that when the users’ walking
rhythm is close to the music rhythm, the users synchronize their gait pattern to
the music’s tempo by taking one step per beat. In a more recent study Leman et
al., (2013) found that participants who had already synchronized their gait tempi
to musical feedbacks increased their speed when they were provided with a more
activating music and decreased it with calm music. These results suggest a
relationship between music type and vigor of the gait.

While the main focus of the above described studies is the effects of musical
stimuli on motor behavior, there are also past studies which seek to investigate
the relationship between either performance or non-performance related auditory
feedback and gait behavior.

More specifically, Menzer et al. (2010) found that inducing delays of different
lengths, ranging from 16ms-1800ms, in the provision of footstep auditory
feedback changed participants’ perception of the exact timing of when they took
each footstep and participants’ speed without them being aware of this change.
The specific research provides evidence that the provision of performance
related auditory cues can affect both motor behavior and perception of body
movement.

In Nordahl’s (2006) study participants were provided with feedback of different

modalities (visual and auditory) during navigation in Virtual Reality

14
environments. The research results showed that the provision of both self-
induced walking sounds which were synthesized in real-time from participants’
walking pattern and of 3D environmental sounds improved the participants’
motion by enhancing their whole body movement rate compared to visual
feedback. However the effects of standalone ego walking sounds did not differ
significantly from the effects of visual feedback.

Zanotto et al. (2013) in their study explored whether feedback perceived by

different sensory channels, such as haptic and visual feedback, could either be
supplemented or substituted by acoustic feedback (rhythmic beats or auditory
feedback synthesized based on the hip and knee flexion). Participants were
asked to walk wearing a robotic exoskeleton and to adopt a new walking pattern
formed based on each participant’s original gait. Different types of guidance
were provided to the participants. The results indicated that combining visual
and kinetic guidance proved to be equally effective with the provision of
combined kinetic and auditory guidance. Additionally, the provision of kinetic-
auditory feedback led to the improvement of gait symmetry and maintenance of
a decreased step height after training. However, complementing kinetic and
visual feedback with acoustic feedback did not improve substantially the
participants’ performance.

In another study Bresin et al. (2010) developed and evaluated a prototype of

interactive sensored-shoes which produced auditory feedback of the sound of
four different ground textures (mud, wood, linoleum and iced snow). Pressure
sensors were used to collect behavioral data. The participants were asked to walk
in four different emotional intentions (happy, aggressive, tender and sad) while
they were provided with the acoustic feedback of the different textures. The
research findings show that participants altered both their motor behavior and
emotions according to the various types of auditory feedback. More specifically,
participants increased their pace when they were provided with an auditory
feedback with a higher spectral centroid sound (e.g. iced snow) and accordingly
decreased their pace when the spectral centroid sound was lower (e.g. mud).
Additionally, participants were found to be more aggressive when they perceived
to walk in harder textures and more tender and sad when the perceived texture

15
was softer. Even though the results are not statistically significant they do
provide evidence of the effects of sound in human motor behavior.

Considering all the above, it is clear that past studies have focused on exploring
the effects of music and other types of auditory cues on gait behavior. While,
Bresin et al. (2010) investigates the effects of real-time walking related sounds
(ground texture sounds) they fact that they asked participants to walk in different
emotional intentions may have confounded the results. In our study we seek to
investigate the effects of real-time self-produced footstep sounds and their
possible influences on walking behavior, emotion and perception.

Sound and Emotions

Emotions play a fundamental role in our everyday lives. Sound is an innate part
of our lives triggering a wide range of emotional reactions and processes during
our daily interactions with the environment. There are past studies revealing the
effects of auditory feedback in emotions (e.g. Bradley and Lang, 2000). In this
section a few researches that investigate the implications of sounds associated to
our body or our actions in human emotions are briefly described.

Tonetto et al., (2014) in their study (details in “Sound and Perception of Body”
section) demonstrated that sounds elicited by high heels of different materials
while walking on different types of ground affect women’s emotions (valence,
arousal and dominance). Additionally, Furfaro et al., (2013) also found effects of
sound in human emotions (valence, arousal, dominance) (details in “Sound and
Perception of Body” section).

Tajadura-Jimenez et al., (2008) investigated the effects of the provision of

auditory and/or vibrotactile heartbeat feedback from different distances
(headphones vs speakerphones) in participants’ emotional state, reaction to a
series of pictures, and memory. The heartbeat feedback was provided to the
listeners in two different rates. The results provide evidence that self-related
auditory feedback affects people’s emotional responses to visual stimuli.

16
Additionally, there is an indication that the provision of auditory stimuli closer to
our body (through headphones) can be more arousing.

While in all the above mentioned studies the auditory stimuli provided to the
participants were pre-recorded, past studies haven’t yet investigated the effects
of real-time, self-produced footstep sounds in human affective states.

Implications for HCI

As showed above footstep sounds constitute an informative source and can have
effects in body perception, motor behavior and emotions. Those three aspects are
highly important for the design of interactive systems. The present research’s
findings can provide valuable insights for the design of applications for virtual
reality, video games, rehabilitation and fitness.

More specifically, Turchet et al. (2010) in their study highlight that footstep
sounds are used in both virtual applications which enhance user navigation and
games to increase the sense of action and resemblance to physical environments.
Additionally, past researches have reported that that “self-representation” sounds
contribute to the development of a virtual self-representation (Väljamäe et al.,
2008) and can be used in VR environments to enhance presence experience
(Tajadura-Jimenez et al., 2008). Thus, the findings of the present study can be
used in the field of virtual reality to enhance presence, immersion and enable a
more realistic interaction in various situations.

Additionally, this study’s findings might be used for rehabilitation systems to

help people with lower limb impairments which are a result of physical or
neurological injuries. For instance, past studies suggested that visual stimuli can
be used to change walking pace in people with stroke (Jaffe et al., 2004).
Furthermore, Rosati et al., (2011) in their study report that the provision of
acoustic feedback (sequences of tonal beeps) during technology-assisted arm
movement in people with stroke enhances patients’ positive emotions. Similarly,
walking generated auditory feedback can benefit people with chronic pain who
show fear and reluctance towards moving (Vlaeyen & Linton, 2000). Singh et

17
al., (2014) in their study revealed that the use of auditory feedback can motivate
physical activity in patients with chronic pain.

Finally, past studies have demonstrated that music feedback during exercise
decreases perceived physical effort and improves affective states (bike riding,
Becker et al., 1994; karate performance, Ferguson et al., 1994). Consequently,
the use of self-produced footstep sounds in the design of interactive personal
training applications could have similar results.

18
2.2 Research Question and Hypotheses
In light of the above mentioned it is clear that previous studies show evidence
that sound related to both our bodies and our actions may influence the
perception of key body attributes, motor behaviour and emotions. However,
there is lack of sufficient understanding of the effects of real-time manipulated
self-produced sounds on these three aspects. Thus, the research question of this
study is:

Can we alter the perception of one’s body weight, behavior and emotions by
manipulating self-produced footsteps sounds?

The specific hypotheses are presented below:

Hypothesis 1

1. Participants’ perception of their own body weight will change with two
possible outcomes (Li et al., 1991):

1.1 Participants will perceive their own body weight as higher when the
low frequency components of their footsteps sound increase.
1.2 Participants will perceive their own body weight as lower when the
high frequency components of their footsteps sound increase.

Hypothesis 2

2. Participants’ motor behaviour will be affected (“Sound and Motor

Behaviour” section) by the provision of manipulated footsteps sound:

2.1 Participants will decrease their speed (longer time interval between
heel strike and toe-off events) when we increase the low frequency
components of the footsteps sound (Bresin et al., 2010). It is also
expected that both pressure applied on the ground and time of contact
with the ground will be increased. Additionally, the acceleration
while lifting the foot to move forward will be decreased (point-light
walker application, Troje, 2008).

19
 2.2 Participants will increase their speed (shorter time interval between
heel strike and toe-off events) when we increase the high frequency
components of the footsteps sound (Bresin et al., 2010). It is also
expected that both pressure applied on the ground and time of contact
with the ground will be decreased. Additionally, the acceleration
while lifting the foot to move forward will be increased (point-light
walker application, Troje, 2008).

Hypothesis 3
3. Participants’ emotional experience will be affected by the provision of
the manipulated footsteps sound with two possible outcomes (Tonetto et
al., 2014; Tajadura-Jimenez et al.; 2008):

3.1 Participants will feel more negative/unaroused/submissive when we

increase the low frequency components of the footsteps sound.
3.2 Participants will feel more positive/aroused/dominant when we
increase the high frequency components of the footsteps sound.

Hypothesis 4
4. Considering the relationship between body weight and walking posture
(point-light walker, Troje, 2008) we believe that participants’ walking
posture will be affected (Pastore et al., 2008; Tajadura-Jimenez et al.,
2012) by the provision of manipulated footstep sounds with two possible
outcomes:

4.1 Participants will adopt a stooped posture (neck and shoulder

inclination will be increased) when we increase the low frequency
components of the footstep sounds.
4.2 Participants will adopt a straight posture when we increase the high
frequency components of the footstep sounds.

20
 Hypothesis 5
5. Participants’ perception of their strength will be affected by the provision
of manipulated footstep sounds (Furfaro et al., 2013) with two possible
outcomes:

5.1 Participants will feel stronger as they perceive their body as heavier
when we increase the low frequency components of the footsteps
sound. Accordingly, they will feel weaker when we increase the high
frequency components.
5.2 Participants will feel weaker as they will feel unfit to a body that is
perceived as heavier when we increase the low frequency
components of the footsteps sound. Accordingly, they will feel
stronger when we increase the high frequency components.

Hypothesis 6
6. Considering the relationship between body weight and walking posture
(point-light walker, Troje, 2008) we believe that participants’ perception
of their walking posture (straight-stooped/hunched) will be affected by
the provision of manipulated footstep sounds (Tajadura-Jimenez et al.,
2012) with two possible outcomes:

6.1 Participants will perceive their posture as stooped when the low
frequency components of the footsteps sounds increase.
6.2 Participants will perceive their posture as straight when the high
frequency components of the footsteps sounds increase.

Hypothesis 7
7. Based on the findings in the “Sound and Motor Behaviour” section which
suggest that sound affects motor behaviour (speed) (Bresin et al., 2010)
and Menzer et al’s., (2010) study which reveals that sound affects
perceived motor behaviour we also believe that participants’ perceived
speed will be affected by the provision of manipulated footstep sounds
with two possible outcomes:

21
 7.1 Participants’ perceived speed will be decreased when we increase the
low frequency components of the footstep sounds.
7.2 Participants’ perceived speed will be increased when we increase the
high frequency components of the footstep sounds.

Additionally, more measures were introduced to further investigate other bodily

sensations (feet localization, experience of surprising and unexpected or vivid
feelings of body and sound source identification). However, the purpose of this
introduction was rather explorative and no specific hypotheses were made for the
effects of footsteps sound in these sensations.

22
 3 PROTOTYPE AND METHODS

3.1 Prototype
For the purpose of the experiment a system based on both audio equipment and
sensors was set up. Additionally, a pair of sensored shoes prototype was build.
In this section a detailed description of the prototype and an overview of the
system used to carry out the experiment will be provided. Furthermore, the way
the different sensors were fitted to the participant will be described. Finally,
details will be given for the exploratory process followed in order to choose the
appropriate equipment that comprises the system.

Prototype and System Description

For the purpose of the present study a pair of sensored active sandals was build.
The sandals were enhanced with pressure sensors and one accelerometer. More
specifically, 4 Square Force-Sensing resistors (43.7 mm x 88 mm) (2 on each
shoe) were placed in the front and rear part of the sandal’s insole (Figure 1). One
ribbon cable (1.5 meters long) exited each of the shoes and was soldered to a
prototyping board along with 4 resistors (330 Ω). The board was then connected
to an Arduino board. Additionally, a Sparkfun 3-axis Breakout accelerometer
integrated into the prototype was attached to participants’ left ankle (Figure 2).
The accelerometer was directly connected to the Arduino board.

Figure 1: The sensored-shoes prototype

23
Figure 2: The sensored-shoes and the accelerometer as it was attached to participants’ left
ankle

The data gathered from both sensors were digitized through Arduino (Figure 3).
More specifically, 3 different boards were used. An Arduino UNO with an
Electronic Brick Shield V4.0 Pro on the top was used. The Electronic Brick
Shield had 5 buckled analog ports. Since we were aiming on reducing cabling
more components were added so that the data would be wirelessly transferred
from the Arduino to the laptop. This was achieved with the addition of an
Arduino XBee board with an XBee wireless component attached to it. An XBee
Explorer USB unit along with an XBee wireless component were also connected
to the laptop to achieve wireless communication with the Arduino board.
Finally, the Arduino UNO board was connected to a 9V battery for power
supply.

Figure 3: The Arduino board

24
A pair of Core sound binaural lavalier microphones with the frequency response
of 20 – 20.000 Hz (one on each shoe) was also integrated into the shoes to
capture the footstep sounds of the walker (Figure 4). The microphones were
directly connected to an SP-24B Stereo Microphone Preamplifier to amplify the
acoustic signal of the footstep sounds. In order to increase the equipment’s
portability the preamplifier was converted into a portable device by connecting it
to an 8 AA battery box. The amplified signal was thereafter transmitted to a
Behringer’s MINIFBQ FBQ800 Ultra-Compact, Graphic Equalizer with 9
frequency bands (83 Hz, 125 Hz, 250 Hz, 500 Hz, 1 kHz, 2 kHz, 4 kHz, 8 kHz,
16 kHz) and 24 dB dynamic range to manipulate the higher and lower frequency
components of the footsteps sound. The participants were required to walk down
an 8.54 meters corridor for the purpose of the experiment. For this reason, the
equalizer’s cable was extended by 15 meters so that the device would be plugged
into the power supply during the walking trial. The sound output was rendered
through a pair of Sennheiser HDA 300 closed headphones with high passive
ambient noise attenuation. The headphones were directly connected to the
equalizer.

Figure 4: The microphones integrated into the sensored-shoes

A Galvanic Skin Response (GSR) sensor (Affectiva GSR Q-sensor by Affectiva)

was placed in the participants’ wrist to measure their skin conductance. The
sampling rate used was 32 Hz. The data were acquired during the experiment,
and recorded on the Q sensor internal memory. They were transferred to the
laptop after each experimental session. Finally, 10 wireless EMG/accelerometer
sensors by Delsy were attached to the participants’ upper and lower part of their

25
body. The sensors were wirelessly connected to their base which was connected
to the laptop with a USB cable. A detailed scheme of the connection between
the different components of the system is presented in Figure 5.

Figure 5: The scheme illustrates the connection between the different components of the
system

Sensor Setup

The audio equipment (preamplifier and graphic equalizer) and the Arduino
prototype board were placed in a small backpack. The backpack was connected
to the shoe prototype. More specifically, 2 ribbon cables and 2 microphone
cables exit the backpack and were connected to the sandals. The accelerometer
cable exited the backpack and was attached to the participants’ ankle with a
hypo-allergenic tape. The participants were first asked to wear the sandals and
they were then assisted to wear the backpack. The cables were attached to the
participant’s legs with the use of two Velcro straps as shown in Figure 6, to
ensure that they could walk comfortably without limitations. The cable

26
 connecting the headphones to the equalizer exited the upper part of the
backpack.

The 10 wireless EMG/accelerometer sensors were attached to the participants’

body as follows. Four sensors (2 on each lower limb) were placed on Tibialis
Anterior and the Medial Head of Gastrocnemious muscles as it is shown in
Figure 6. Two sensors were placed on the back of the participants’ neck and their
right shoulder (Figure 7). The rest 4 sensors (2 on each lower limb) were placed
on the ankles and on the side of the thigh above the knee (vastus lateralis)
(Figure 8). Finally, the GSR sensor was placed on the participants’ wrist. The
whole setup is illustrated in Figure 10.

Figure 6: The Velcro Figure 7: The Figure 8: EMG/accelerometer

straps used to attach the EMG/accelerometer sensors sensors placed on
cables to participants’ placed on participant’s neck participant’s ankle and the
lower limbs and right shoulder side of their thigh

27
 Figure 9: The GSR sensor placed on participant’s non-dominant hand

Figure 10: Shots of the setup

28
Exploration

The researcher went through an initial exploratory stage to investigate the

functionality and efficiency of different types of sensors and audio equipment to
ensure the final system’s effectiveness.

Shoes

Sandals were preferred to regular shoes since they are easy to fashion and they
can accommodate a wide range of foot sizes. Two different types of sandals were
tested regarding the sound elicited from the sole impact with the ground. A pair
of Arlington sandals by Earth Spirit, UK size 8 (EU, 42) was chosen primarily
due to the hardness of the sole material (hard rubber) which elicited a clearer and
more distinctive footstep sound.

Audio equipment

Regarding the sound input there was a need for microphones of high quality that
could capture a wide range of both the high and low frequency footstep sound
components and could be easily attached to the shoes prototype. For this reason
four different types of microphones were tested to choose the most effective one
(Core sound binaural microphone).

Additionally, the Sennheiser HDA 300 headphones were tested against 2

different types of compact phone speakers. While the phone speakers would
allow us to provide the sound from its original location (shoes) they were not
preferred since they were heavy, they produced a sound output of low quality
and they did not mask environmental sounds. Thus the headphones were
considered to be more appropriate.

Two types of equalizer were initially tested (Altai Soundlab DEQ31X1 Equalizer
and Audacity 2.0.5 virtual equalizer) to ensure that we could achieve the
desirable manipulations of the frequency spectrum. Altai Soundlab DEQ31X1
Equalizer was highly effective. However, its large size didn't allow us to use it
for the project. Thus we purchased an equalizer with similar characteristics and
smaller dimensions. (Behringer’s MINIFBQ FBQ800 Ultra-Compact, 9-Band
Graphic Equalizer).

29
Sensors

Three different types of Force-Sensitive sensors (2 sizes of round Force-

Sensitive sensors and a Square one) were first tested in terms of their ability to
detect micro changes in the pressure exerted from the foot on the ground (Figure
11). While there was no significant difference in their degree of sensitivity to
detect changes in the applied pressure the Square sensor was larger in
dimensions and had better applicability since it could detect the force applied in
a larger area.

Figure 11: testing different types of pressure sensors

Finally, a systematic analysis was performed in order to identify the most

appropriate location for the accelerometer. For this purpose, 3 fellow students
were asked to perform 2 walking trials each. One trial with the accelerometer
attached to their knee and one where the accelerometer was attached to their
ankle. The data gathered were plotted in Microsoft Excel 2010 and a quick
analysis revealed that the range of change in the signal was larger on the ankle.

30
3.2 Methods

Participants

Data were gathered from 22 participants (4 males and 18 females) with ages
ranging from 18 to 35 years (M=24.36; SD=4.85) The participation criteria set
were the following: age between 18 and 35, normal hearing, willing to mostly
stand upright for about an hour, no neurological/psychiatric disorder. All
participants were naive as to the purpose of the study. The participants were
given £7.5 for their time.

Materials and Apparatus

Prototype and Hardware

The prototype was described above.

Additionally, a Dell Desktop computer was used to run a 3D Body Visualiser

application used for the purpose of the experiment. All the data gathered from
the sensors were transmitted via Bluetooth and recorded in a Toshiba Satellite
L750 17R.

Software

The Arduino integrated development environment (IDE) software (Arduino

1.0.5) and Processing development environment were used. The processing
programme was used for both plotting the data in real-time and data recording
(Figure 12). Both the Arduino and processing scripts were provided by the
supervisor and only small alterations were made on the code by the researcher.
The scripts are provided in Appendix A and B.

31
Figure 12: Real-time plotting of data in Processing environment – the first four are the
pressure sensor signals. The last three are the plots of the accelerometer x,y,z axes.

Additionally, the Q-Sensor software (Figure 13) was used to acquire, plot and
export the GSR data to .csv files so that they could be utilized for further
analysis on Matlab.

Figure 13: Q-software interface – plot of the GSR signal of participant 1

Delsy’s EMGworks 4.0 Acquisition software was installed for the acquisition of
Delsy’s EMG/accelerometer sensor data (Figure 14).

32
Figure 14: Delsy’s EMGworks 4.0 Acquisition software – plot of electromyographic
activation and accelerometer data

Finally, a 3D Body Visualization application was used to measure participants’

perceived body weight after each experimental trial (http://bodyvisualizer.com/).
The applications layout consisted of a 3D body avatar located on the left of the
screen and the values of the body measurements on the right of the screen. A
paper board was used to cover the right part of the screen (Figure 15) so that the
measurements indicators were not visually accessible by participants. Thus, bias
derived from seeing the dimension values was eliminated. The specific measure
allows changing the avatar body across different dimensions. For the study’s
purpose the experimenter fixed the height dimension to match each participant’s
own height, and then selected the weight parameters, so that participants would
only make changes in this dimension. The avatar’s weight could be manipulated
by using the right and left arrows of the keyboards (more details in “Measures”
section below).

33
 Figure 15: The 3D Body Visualizing application

Sound manipulation

As mentioned above the footsteps sound was amplified by a stereo microphone

preamplifier before being transferred to the equalizer. As a result of the
amplification the overall sound output was attenuated by 12 dB to a more
comfortable listening level.

The sound manipulation was guided by the design in Li et al’s., (1991)

experiment. However, we made small modifications on the frequency ranges
based on our own listening experience.

More specifically, there were 3 different sound conditions. A neutral condition in

which participants were provided with their natural footsteps sound amplified
(Figure 16). A High Frequency condition in which the high frequency
components of the footsteps sound in the range between 1 kHz – 4 kHz were
amplified by 12 dB, while the low frequency components in the range between
83 Hz – 250 Hz were attenuated by 12 dB (Figure 17). The Low Frequency
condition in which the low frequency components of the footsteps sound in the
range between 83 Hz – 250 Hz were amplified by 12 dB, while the high
frequency components above 1 kHz were attenuated by 12 dB (Figure 18).

34
Figure 16: Settings for the Figure 17: Manipulation Figure 18: Manipulation of the
neutral condition – no of the high frequency low frequency components
frequency manipulation components

Measures

Quantitative data were gathered through a 3D Body Visualization application, 3

questionnaires, 4 Force-Sensitive resistors, an accelerometer, a GSR sensor and
10 wireless EMG/accelerometer sensors. The measures are described below. All
questionnaires are presented in Appendix C.

3D Body Visualization Application

The 3D body Visualization application was utilized to measure participants’

perceived body weight after each experimental trial. The participants were asked
to adjust the weight of the body avatar to correspond to the way they perceived
their own body weight at the specific time. The default settings of the avatars
weight were adjusted by the experimenter before each trial to either ±25% of the
participant’s actual weight. The default settings were randomized across the 2
repetitions of each condition.

Questionnaires

Questionnaire 1

A preliminary questionnaire regarding participants’ age, weight, height and shoe

size was administered (See Questionnaire 1 in Appendix C). Furthermore, more
questions regarding participants habits such music education and exercise habits
were introduced to the questionnaire to mislead participants about the aim of the
study and thus to avoid introduction of bias in the results.

35
Questionnaire 2 - Spanner questionnaire

A fake task (spanner task) was introduced to the experiment for two reasons: To
mislead the participants regarding the actual purpose of the study and to reduce
visual contributions to the task that could lead to biased responses, meaning that
participants would focus on the task and would not inspect their gait patterns
during the experimental trial. The participants were presented with three
spanners of different sizes (Figure 19) and were informed that they would be
given one of these spanners before each experimental trial. However, they were
given the same spanner in each trial. The participants were asked to make
approximate assessments of the spanner’s length and weight. Despite this being a
fake task the participants’ responses were later analyzed to detect possible
effects of sound on their assessments. (Full questionnaire in Appendix C-
Spanner Questionnaire).

Figure 19: The 3 different spanners presented to the participants

Questionnaire 2 - Perceived Self efficacy Questionnaire

Perceived self-efficacy is defined as a person’s ability to carry out a specific task

and its degree varies across people and conditions (Bandura, 2006). Bandura
(1977) in his study presents a theory suggesting that perceived self-efficacy can
be changed by any kind of psychological procedure. In 2006 Bandura developed
a series of different scales for measuring self-efficacy in different domains such
as the Self-Efficacy scale to regulate eating habits. For the purpose of the present
study the Self-Efficacy scale for lifting objects of different weights was used to

36
further explore whether the different sound manipulations affect participants’
perceived strength. (Full questionnaire in Appendix C-Lifting Questionnaire).

Questionnaire 2 - Self-Assessment Manikin (SAM)

Participants’ emotional valence, arousal and dominance levels related to the task
were measured by using the Self-Assessment Manikin (SAM), a 9-point non-
verbal pictorial assessment method introduced by Bradley and Lang (1994). The
specific method is used to measure the aforementioned affective responses in a
variety of stimuli including sound. Since SAM is a non-verbal method it can be
used across different cultures and it quick to fill in and highly reliable despite its
simplicity (Bradley and Lang, 1994). (Full questionnaire in Appendix C-
Questionnaire 2.2).

Questionnaire 2 - Task experience questionnaire

7-point likert scales were used to measure the way participants perceived their
walking speed, weight, strength and posture after each experimental trial.
Additionally, the same type of scales was used to measure participants’ degree of
agreement regarding aspects such as the feelings of their body and their ability to
locate their feet. In this case higher values corresponded to higher level of
agreement with each question. (Full questionnaire in Appendix C-Questionnaire
2.1).

Questionnaire 3-Eating Disorder Examination Questionnaire (EDE -Q)

The EDE-Q is the self-report version of the Eating Disorder Interview (EDE)
introduced by Fairburn and Cooper (1993). The weight and shape concern sub
scale items of the specific questionnaire were utilized to explore the participants’
degree of concern regarding their weight and shape and identify possible
correlations between the level of shape/weight concern and manipulation of body
perception. (Full questionnaire in Appendix C-Questionnaire 3).

37
Pressure sensors

The Force-Sensitive resistors were utilized to measure the interaction force

(pressure applied on the ground) from both the heel and the ball of the foot and
participants’ pace.

Accelerometer

The accelerometer was used to measure the acceleration and deceleration of

participants’ left foot movements during the Swing phase in x, y and z axis.

Galvanic Skin Response

Galvanic Skin Response (GSR) measures the physiological changes in skin

electric conductance which occur due to reaction to psychological events
(Montagu & Coles, 1966). GSR is largely applied to detect people’s reactions in
visual or acoustic feedback (Lunn and Harper, 2010). The specific measure has
been used in past studies to measure affective states such as frustration (Lunn
and Harper, 2010), stress level (Perala and Sterling, 2007) and arousal (Lang,
1995). However, it has been heavily criticized regarding its limitation to
accurately detect the specific psychological event that occurs (Lunn and Harper,
2010). Thus, using GSR in combination with other measures has been suggested
(Lunn and Harper, 2010).

EMG/accelerometer sensors

Delsy’s wireless sensors were used to measure the electromyographic activity of

Tibialis Anterior and the Medial Head of Gastrocnemius muscles and the
inclination of neck, right shoulder, shank and thigh.

Other materials/Environment/Lighting

The experimental sessions were recorded with a camera (JVC Everio). In order
to minimize visual distractions the lighting of the room was lowered during the
experiment. Finally, 4 MDF boards with dimensions 2440mm x 1220mm x
25mm where placed next to each other forming an 8.54 meters corridor (Figure
20).

38
 Figure 20: The experiment setting

Consent Form

An Informed Interviewee Consent Form was signed by the participants before the
beginning of the experiment (See consent form in Appendix D)

Design

The experimental design was a 3x2 within participants factors. The independent
variables were the 3 different types of auditory feedback provided to the users as
described above (veridical feedback, high frequencies amplified and low frequencies
amplified). Two repetitions were performed for each type of auditory feedback.
Consequently, 6 experimental trials were performed by each participant. The
provision of the auditory feedback provided in each trial for each participant was
randomly ordered. The dependent variables were the perceived body
weight/dimensions, assessment of spanner length and weight, self-efficacy regarding
lifting objects, the task experience questions, perceived valence, arousal and
dominance, GSR, acceleration and deceleration of foot movement, pressure applied
from the heel and ball of foot to the floor and walking pace.

39
Procedure

The experiment lasted approximately one hour. It took place at the Institute of
Philosophy, Senate House, in London. Prior to the experiment participants were
informed that they should wear a pair of shorts and socks. At the moment the
participants arrived in the experiment room, they were asked to read through an
information sheet (Appendix H) with a detailed description of the experimental
procedure and to sign the Informed Interviewee Consent Form. They were also
required to fill in the preliminary questionnaire concerning demographic and
personal information. Thereafter, the GSR sensor was placed on the participant’s
non-dominant hand. After showing the different spanners to the participant the
researcher role-played the whole experimental task. Two initial practice blocks
of 2 trials (one without the equipment on and one with the equipment on) were
performed prior to the experimental block to allow participants to familiarize
with both the equipment and the task. Before the beginning of the actual
experimental sessions 2 small tests were performed to collect the maximum
activation values of the Tibialis Anterior and the Gastrocnemious Medial Head
muscles. These values were essential for the normalization of the
electromyographic data that would be collected subsequently. Test 1 required
participants to lift their heels 5 times to collect the maximum activation values of
Tibialis Anterior. For the collection of the maximum values of Gastrocnemius
Medial Head participants were first asked to lift their toes. Their toes were then
pushed towards the floor and the participant would have to resist the pushing.
Once the tests were carried out the experiment began. Participants were provided
with the spanner. A “click and start” signal was then given to participants to
press the GSR sensor’s button which set a marker in the data and start marching
at place. The purpose of the introduction of the marching task was to achieve
longer exposure the sound stimuli. After 10 seconds the researcher gave a “go”
signal and the participants started walking towards the end of the corridor where
they were required to press the GSR button again, put the spanner in a non-
transparent bag, adjust the measurements of the 3D body avatar and fill in
Questionnaire 2 described above (“Measures” section). Participants were asked
to walk at their comfortable speed. Subsequently a new session started. This exact
procedure was performed 6 times. After the end of the 6 experimental trials the

40
researcher helped the participant to take the equipment off. Finally, the EDE-Q
questionnaire was filled in and the participant was debriefed and paid.

41
3.3 Data Extraction
MATLAB scripts were used to extract the final data for the subsequent analysis.
Please note that the scripts were developed with the supervisor’s help. The
scripts can be found in Appendix E.

Behavioral and Physiological Data

Force Sensitive Sensors

The data captured by the pressure sensors were integer numbers of different
values. This variation illustrated the different levels of pressure applied on the
floor across time. The values were plotted in MATLAB as illustrated in Figure
21. Thereafter, the start and end points of the walking session were manually
identified for each of the 4 pressure sensors and the signal was cut. Since the
marching part was mainly introduced to increase sound exposure, it was not
included in the analysis. The differentiation between the pressure values given
during the marching and the walking session is obvious in Figure 21. Five main
variables were extracted from the heel and ball of foot for each step as presented
in Table 1. Figure 22 illustrates the main variable in a plot of the pressure signal.
Subsequently, 7 seven final variables were calculated and used for the analysis to
detect significant changes in participants’ pace and/or pressure applied on the
floor. The variables extracted to measure participants pace were: the duration of
heel and toe contact with the ground and the time interval between the heel strike
and toe-off events. The pressure applied on the ground was measured by
analyzing the maximum and average pressure applied on the ground from both
heel and ball of foot (toe area). The final variables are illustrated in table 2.

42
Figure 21: A plot of the signal extracted from the left toe pressure sensor. Both the
marching and walking session are illustrated.

Variable label Explanation

t0 The time of the heel/toe strike event –
the minimum value below a specific
threshold, between 2 peaks
t1 The time of the heel-off/toe-off event -
– the minimum value below a specific
threshold, between 2 peaks
p The maximum heel/toe pressure value -
maximum values were the peaks
identified above a specific threshold.
tp The time of the heel/toe maximum
value
avg The average heel/toe pressure
Table 1: The 5 main variables extracted from the heel and ball of foot

43
 p,tp

avg

t0
t1

Figure 22: A plot of the left toe pressure signal during the walking session and the main
variables extracted from the heel and ball of the foot

Variable name Explanation

heel_t1t0diff Duration of heel contact with the
ground
heel_to_toe Time interval between heel strike and
toe-off events
heel_peak Maximum heel pressure applied on the
ground
heel_average The average heel pressure applied on
the ground
toe_t1t0diff Duration of toe contact with the
ground
toe_peak Maximum toe pressure applied on the
ground
toe_average The average toe pressure applied on
the ground
Table 2: The final variables that were calculated and utilized for the analysis

44
Accelerometer

The data captured by the accelerometer were a string of positive and negative
numbers. These variations suggest the accelerating and decelerating movements of
the foot in x,y and z axes during the Swing phase. While acceleration data were
gathered from all 3 axes the resultant of the 3-axes data was calculated as it follows
to facilitate subsequent analysis:

Concerning the definition of the end and start point of the walking session the
same process as the one described above was followed. Figure 23 shows an
acceleration and pressure plot before cutting the signal according to the start/end
points of the walking phase. A plot of the acceleration during the walking trial is
showed in Figure 24. After plotting the data maximum and minimum values
above and below specific thresholds were identified. Thereafter, 3 final values
were calculated and used for subsequent analysis to detect changes in
participants’ acceleration of movements as showed in table 3

Acceleration values Definition

acc_lifting The maximum acceleration value when
the foot after the toe-off event, when
the foot leaves the ground to move
forward.
dec_lifting The maximum deceleration captured
during the ascending foot movement
when the foot prepares to descend
towards the ground
acc_down The maximum acceleration value
captured during the descending
movement of the foot towards the
ground
Table 3: Accelerometer final variables

45
Figure 23: Plot of acceleration and pressure recorded from all 4 pressure sensors

Figure 24: Acceleration during walking session

GSR

A MATLAB script (Appendix F) was used to identify the changes in the GSR signal
during the walking session. Before exporting the data on MATLAB the GSR files
were plotted in Q-Sensor software and extra markers were added for the
differentiation of the walking and the marching session. Data were then imported
into MATLAB and values were calculated for subsequent analysis (Table 4)

Measures Definition
avg_march Average arousal during marching

46
avg_walk Average arousal during walking
max_min_march Difference between maximum and
minimum arousal during marching
max_min_walk Difference between maximum and
minimum arousal during walking
Table 4: GSR main variables

Participants were asked to wear the sensor during the whole experimental
procedure. Thus one GSR file was exported for each participant. In order to
identify the data relevant to each experimental trial the participants were asked to
press the sensor’s button in both the beginning and the end of each trial so that
markers would be placed on the data files.

47
 4 DATA ANALYSIS

Data extraction was followed by the statistical analysis using the software IBM
SPSS 22. Due to time limitations, the analysis of the data gathered using the
DELSY’S Trigno EMG System and the EDE-Q was postponed to future
analysis. For the purpose of this report four types of data were analysed, the
questionnaire, the pressure sensors, the accelerometer and the GSR data.

For all the variables, initially exploratory analyses were performed to test
whether the distribution as a whole deviates from a comparable normal
distribution. The objective Shapiro-Wilk test of normality was used since it is
reported to be more accurate than the Kolmogorov-Smirnov one, (Field, 2005).

In the case of a non-normal distribution, four different transformations were

tested to achieve data normality in the order presented below. After each
transformation a normality test was run and no more transformations were
performed if normality was achieved.

o The log transformation (log(Xi)): takes the logarithms of the initial data.
In the case of zero or negative values, log(Xi+1) was calculated.
o The square root transformation (√Xi): takes the square roots of each of
the scores. In the case of negative values √(Xi+1) was calculated.
o Reciprocal Transformation (1/Xi): divides 1 to each score of the initial
data.
o In the case of zero values 1/(Xi+1) was calculated.
o Z-scores: the initial data are converted to individual z scores calculated
for each participant based on his/her data gathered from 6 trials according

to the following equation:

In certain cases (avatar task, perceived weight, average time of heel contact,
acceleration when lifting foot) and further examination to detect outlier data was
conducted, by looking for values that exceeded or were under two SD from the
mean. However, no outlier data were detected in any of the cases.

48
According to the results of the normality test either parametric or non-parametric
analysis was carried out, as described below.

In the case of a normal data distribution, repeated measures analysis of variants

(3 x 2 ANOVAs, sound by repetition) or multivariate analysis of variance
(MANOVA) were conducted to identify significant sound effects and
interactions. When there was no indication of significant effects of sound
repetition the data of the two repetitions for each sound condition were merged
and the Mean Values and Standard Errors were calculated. Thus, the Mean
values were used and subsequent analysis was performed in three levels instead
of six. Thereafter, planned pairwise comparisons within the three sound
conditions were performed through paired sample t-tests to further explore
possible trends or significant sound effects between specific conditions. Finally,
the effect size (r) in paired sample t-tests was calculated as follows:

, t= t value, df= degrees of freedom

In the case of a non-normal data distribution, the two repetitions of each sound
condition were compared against each other by conducting a Wilcoxon test, to
detect possible significant effects of sound repetition. In the case of no
significant sound repetition effects the Means and SD were calculated.
Thereafter, Friedman’s test analysis was performed to detect whether there is a
significant sound effect within the different sound conditions. Finally, a
Wilcoxon test was conducted to detect significant sound effects and trends
within specific conditions. The effect size in Wilcoxon test is demonstrated by
reporting r. The value was calculated as follows:

, z= z-score, N= number of total observations

Results and analysis are reported specifically in the next sections of the chapter.
There is a separate section for each data type (questionnaire, pressure sensors,
accelerometer and GSR sensor). All the detailed tables with the statistical test
outputs are included in Appendix G.

49
Sound condition Labelling
No sound manipulation – repetition 1 NF1
No sound manipulation – repetition 2 NF2
Manipulation of High Frequencies- HF1
repetition 1
Manipulation of High Frequencies – HF2
repetition 2
Manipulation of Low Frequencies - LF1
repetition 1
Manipulation of Low Frequencies – LF2
repetition 2
Table 5: Conditions labelling

Sound Condition Labelling

No sound manipulation – mean value NF_mean
of two repetitions
Manipulation of High Frequencies - HF_mean
mean value of two repetitions
Manipulation of Low Frequencies - LF_mean
mean value of two repetitions
Table 6: Labelling for the mean values of the two repetitions for each condition

4.1 Questionnaires
Different types of questionnaires were used for the purpose of the experiment.
Due to time limitations, the Eating Disorder Examination Questionnaire (EDE-
Q) (Fairburn, 2008) was excluded from the current analysis.

Perceived body weight as measured by the 3D Body Visualization

Application

The initial Shapiro-Wilk test revealed that the data distribution was significantly
non-normal, p<.05. Normalisation was achieved by performing a Reciprocal

50
transformation (1/Xi) on the initial data. A sound x repetition ANOVA was
conducted to explore possible sound effects and interaction. Mauchly’s test
showed no violation of sphericity (2)=5.46, p=.065. Thus we are allowed to
report the sphericity assumption, F(2,42)=3.58, p=.037 which showed significant
sound effects on perceived body weight. No effects of sound repetition or
interaction were found. The t-test revealed that participants perceived their own
body weight as lighter in HF (M = 54.65,SE = 2.72) than in NF condition (M =
57.52, SE = 3.02), t(21)=-2.4, p=.025, r=.46 . Additionally, a trend to significant
sound effects was also revealed in HF with respect to LF condition (M = 56.77,
SE = 2.88), t(21)=-1.77, p=.090<0.1, r=.36. There was no significant difference
in LF with respect to NF, p>.05.

Perceived Body Weight

62
60
58
56 No_Frequency_mean

Kg 54 Low_Frequency_mean
52 High_Frequency_mean

50
48
46

Figure 25: Means and Standard Errors for perceived body weight in each sound condition
– the non-transformed initial values are reported

Task Experience Questions

The Task Experience Questionnaire was comprised of eight questions.

Significant effects of sound were found in the values of five measures, p<.05. No
significant effects were found for the rest of them. In seven out of eight measures
data were non-normally distributed and normalisation could not be achieved.
Detailed results for each measure are demonstrated below.

51
Perceived speed - Slow-Quick scale

The Shapiro-Wilk test showed significantly non-normal data distribution (all

p<.05). No significant effects of sound repetition were revealed in the
subsequent Wilcoxon test, p>.05. Consequently, the data of the two repetitions
for each condition were merged and their mean values (Figure 26) were utilised
for the following Friedman’s test. The Freidman’s results suggested a significant
sound effects in perceived walking speed, (2)=8.36, p=.015<0.05. A
Wilcoxon test within the three different sound conditions revealed that perceived
walking speed was significantly higher in HF (M = 5.06, SE = .21) than in LF (M
= 4.13,SE = .25) z=-2.54, p=.011, r=-.38 and NF conditions (M = 4.31,SE = .25),
z=-2.03, p=.042, r=-.30. There was no significant effect of sound in LF with
respect to NF condition, p>.05. The results suggest that participants felt quicker
in HF condition.

Perceived Speed
6

4 NF_slow_quick_mean
3 LF_slow_quick_mean

2 HF_slow_quick_mean

Figure 26: Means and Standard errors for perceived walking speed values in each sound
condition - for the measurement an 7-point LIkert scale was utilised, where 1 was slow and
7 was quick

Perceived weight – Light-Heavy scale

The Shapiro-Wilk test showed non-normal data distribution (p<.05).

Furthermore, a Wilcoxon test showed no significant effect of sound repetition
(p>.05). Friedman’s test demonstrated no significant effects of sound in
perceived body weight, (2)=4.20, p=.122>0.05. However, the fact that p value

52
in Friedman’s test was not far from significance indicated that possible sound
effects or trends could be revealed by a further Wilcoxon test between the three
sound conditions. The results showed that participants perceived their own body
weight as lower in HF (M = 3.65,SE =.21) than in LF condition (M = 4.40,SE
=.24) z=-2.02, p=.043, r=-.30. Additionally, there was a trend to significance in
HF with respect to NF (M = 5.29,SE = 1.19), z=-1.71, p=.087< 0.1, r=-.25. No
significant effects of sound were detected in LF with respect to NF condition,
p>.05. The results suggest that participants felt lighter in HF condition and
heavier in NF condition (Figure 27). A variation in perceived body weight across
conditions is also demonstrated. These results are consistent with the ones
yielded previously in the analysis of data regarding the perceived body weight.

Perceived Body Weight

7

5
NF_light_heavy_mean
4
LF_light_heavy_mean
3
HF_light_heavy_mean
2

Figure 27: Means and Standard errors for perceived body weight in each sound condition -
for the measurement a 7-point Likert scale was utilised, where 1 was light and 7 was heavy

Perceived strength- Weak-Strong scale

The Shapiro-Wilk test showed that data were non-normally distributed (p<.05)
and the consequent Wilcoxon test revealed no significant effects of sound
repetition, p>.05. Friedman’s test showed that sound effects were insignificant
but not far from significance, (2)=3.75, p=.153. Hence, paired comparisons
were planned based on our hypothesis. A Wilcoxon test was run to detect
possible significant effects and trends. The results demonstrated a trend to

53
significant difference in participants’ perceived strength in HF (M = 4.59,SE
=.14) with respect to the LF condition (M = 4.11,SE =.18) z=-1.80, p=.071, r=-
.27. No statistically significant difference was found in the rest paired
comparisons (all p>.05).

Perceived Strength
5

4
NF_weak_strong_mean
3
LF_weak_strong_mean
2
HF_weak_strong_mean
1

Figure 28: Means and Standard Errors for perceived strength values in each sound
condition - for the measurement a 7-point LIkert scale was utilised, where 1 was weak and
7 was strong

Perceived posture - Stooped/Hatched – Straight scale

Data distribution was non-normal. A Wilcoxon test revealed a significant effect

of sound repetition between HF1 and HF2 conditions, z=-2.003, p=.045. Thus,
the data of the repetitions of each condition could not be merged and subsequent
Friedman’s test and Wilcoxon test where run separately for the sound conditions
in repetition 1 and in repetition 2. The Means and Standard Errors of the
perceived posture values for the two different repetitions are demonstrated in
Figure 29 and Figure 30 respectively. Friedman’s test showed no significant
sound effects in both repetition 1, (2)=2.0, p=.357 and 2, (2)=3.83, p=.147.
However, the p value in repetition 2 along with our hypothesis, suggested that
possible trends or significant effects of sounds may be revealed by a further
Wilcoxon test. Thus paired comparisons were planned. The results showed a
significant effect of sound in perceived posture in HF2 (M = 5.59,SE = .28) with
respect to NF2 condition (M = 4.86,SE = .27) z=-2.09, p=.036, r=-.44. No
54
significant effects were found in the rest compared pairs of sound conditions (all
p>.05).

Perceived posture
6

5
NF_stooped_hatched_straig
4 ht
LF_stooped_hatched_straig
3
ht
2 HF_stooped_hatched_straig
ht
1

Figure 29: Means and Standard Errors for perceived posture values in repetition 1 - for
the measurement a 7-point Likert scale was utilised, where 1 was stooped/hunched and 7
was straight

Perceived Posture
7
6
NF2_stooped_hatched_strai
5
ght
4 LF2_stooped_hatched_strai
3 ght
HF2_stooped_hatched_strai
2
ght
1
0

Figure 30: Means and Standard Errors for perceived posture values in repetition 2 - for
the measurement a 7-point Likert scale was utilised, where 1 was stooped/hunched and 7
was straight

55
Feet localization

The data were non-normally distributed (all p<.05). The Wilcoxon test revealed
no significant effects of sound repetition (all p>.05). A significant effect of
sound in the ability of participant to identify the location of their own feet was
found in Friedman’s test, (2)=8.51, p=.014<0.05. A further Wilcoxon test was
run to identify significant effects and trends within pairs of conditions. The
results suggested that participants appeared more confident to identify their feet
location in HF (M = 5.81,SE =.21) than in LF (M = 5.25,SE = .29) z=-2.06,
p=.039, r=-.31 and NF conditions (M = 5.22, SE = .28) z=-2.29, p=.022, r=-.34.
No significant effects of sound were identified in LF with respect to NF
conditions (p>.05).

Feet Localization
7
6
NF_Identification_of_foot_l
5
ocation_mean
4 LF_Identification_of_foot_lo
3 cation_mean
HF_Identification_of_foot_l
2
ocation_mean
1
0

Figure 31: Means and Standard Errors for the values of participant’s ability to identify
their feet location - for the measurement a 7-point LIkert scale was utilised, where 1 was “I
strongly agree” and 7 was “I strongly disagree” with the statement: During the experience I
could really tell where my feet were

Participants’ ability to identify their own feet as the source of footstep sounds

The data distribution was non-normal (all p<.05). Wilcoxon test showed no
significant effect of sound repetition, (all p>.05). There were no significant
effects of sound in participants’ ability to identify their own feet as the source of

56
the footstep sounds, (2)=.241, p=.88. Mean values and Standards Errors are
demonstrated in Figure 32.

Sound source identification

7

6.5
NF_sounds_source_mean
6 LF_sound_source_mean
HF_sound_source_mean
5.5

Figure 32: Means and Standard Errors for participants’ ability to identify their
own feet as the sound source of footstep sounds - for the measurement a 7-point LIkert
scale was utilised, where 1 was “I strongly agree” and 7 was “I strongly disagree” with the
statement: During the experience I felt the sounds I heard were produced by my own
footsteps/ body

Experience of surprising and unexpected feelings of body

Shapiro-Wilk test showed a normal data distribution. No effects of sound

repetition were identified in the Wilcoxon test. Additionally, the sound x
repetition ANOVA test showed no significant sound effects or interaction (all
p>.05). Mean values and Standards Errors are demonstrated in Figure 33.

Surprising and Unexpected Fellings of Body

5

4 NF_surprising.unexpected
_feelings_mean
3 LF_surprising.unexpected
2 _feelings_mean
HF_surprising.unexpected
1 _feelings_mean
0

Figure 33: Means and Standard Errors for the experience of surprising and unexpected
feelings of body - for the measurement a 7-point LIkert scale was utilised, where 1 was “I

57
strongly agree” and 7 was “I strongly disagree” with the statement: During the experience
the feelings about my body were surprising and unexpected

Vividness of perceived body feelings

The data distribution was non-normal (all p<.05) and no effect of sound
repetition was found (all p>.05). A Friedman’s test that was performed showed
no significant sound effects or interaction (all p>.05). Means and Standard
Errors are demonstrated in Figure 34.

Vividness of perceived body feelings

4
3.5
NF_vividness_of_body_feel
3
ings_mean
2.5
LF_vividness_of_body_feeli
2
ngs_mean
1.5
HF_vividness_of_body_feel
1
ings_mean
0.5
0

Figure 34: Means and Standard Errors for the degree of vividness of perceived body
feelings in each condition - for the measurement a 7-point LIkert scale was utilised, where 1
was “I strongly agree” and 7 was “I strongly disagree” with the statement: During the
experience the feeling of my body was less vivid than normal

Self-Assessment Manikin (SAM)

The Shapiro-Wilk test was conducted for all the measures in order to identify the
normality of the data distribution. All the variables were significantly non-
normal and normalisation could not be achieved.

SAM – Dominance

The Wilcoxon test showed no significant effect of sound repetition, p>.05.

Additionally, there were no significant sound effects in dominance,
(2)=2.47,p=.29. However, the mean values of dominance (Figure 35) across

58
different conditions indicated that possible significant sound effects or trends
could be identified. A further Wilcoxon test between the three different sound
conditions showed that participants felt significantly more dominant in HF
(M=6.02,SE=.26) than in LF condition (M=5.29,SE=.32) z=-2.00, p=.045. A
trend is identified in HF with respect to NF condition (M=5.27, SE=.29), z=-
1.74, p=.080. No statistically significant sound effect was found between LF and
NF condition, p>.05.

SAM-Dominance
7

5 NF_SAM_Dominance_me
an
4
LF_mean_SAM_Dominanc
e_mean
3
HF_mean_SAM_Dominan
2 ce_mean

Figure 35: Means and Standard Errors for perceived dominance in each condition - for the
measurement a 7-point LIkert scale was utilised, where 1 was Submissive/Awed and 7 was
Dominant/Important

SAM – Valence

No significant effect of sound repetition was detected (all p>.05). Friedman’s

test indicated a significant effect of sound in valence, (2)=6.22,p=.045. The
Wilcoxon test showed that participants felt happier and more positive in HF
(M=6.29,SE=.23) than in NF condition (M=5.45,SE=.23) z=-2.39, p=.017, r=-
.36. A trend was found in HF with respect to LF condition (M=5.52,SE=.26) z=-
1.81, p=.069, r=-.27. No statistically significant difference was found between
the rest paired conditions, p>.05.

59
 SAM - Valence
7
6
5
NF _SAM_Valence_mean
4
LF_SAM_Valence_mean
3
HF_SAM_Valence_mean
2
1
0

Figure 36: Means and Standard Errors for perceived valence in each condition -
for the measurement a 7-point LIkert scale was utilised, where 1 was Unhappy/Negative
and 7 was Happy/Positive

SAM – Arousal

No effect of sound repetition was detected in Wilcoxon test, (all p>.05).

Friedman’s test showed no significant effects of sound in perceived arousal,
(2)=2.32, p=.31. However, based on our hypothesis, a Wilcoxon test was
carried out. A trend was found in HF (M=5.00, SE=.28) with respect to NF
condition (M=4.40,SE=.32) z=-1.92, p=.055, r=-.28. No statistically significant
difference was found between the rest paired conditions, p>.05.

SAM - Arousal
6

4 NF_SAM_Arousal_mean
3 LF_SAM_Arousal_mean

2 HF_SAM_Arousal_mean

Figure 37: Means and Standard Errors for perceived valence values in each condition - for
the measurement a 7-point LIkert scale was utilised, where 1 was Unaroused/Calm and 7
was Aroused/Excited

60
Spanner Questionnaire

The initial Shapiro-Wilk test revealed that the data distribution was significantly
non-normal, p<.05. The initial data were converted to z-scores to achieve
normalisation. A multivariate analysis of variance (MANOVA) was conducted
to detect sound effects, interaction and possible correlations between the weight
and length variables. The results suggested no significant effects or interaction
(all p>.05). Figure 38, 39 illustrate the means and Standard Errors for perceived
spanner length and weight respectively.

Perceived Spanner Weight

500

400

300 NF_spanner_Weight
Gram (g) LF_spanner_Weight
200 HF_spanner_Weight

100

Figure 38: Means and Standard Errors for perceived spanner weight values in each
condition

Perceived Spanner Length

20

19

NF_spanner_Length
18
Cm LF_spanner_ Length
17
HF_spanner_ Length

16

15

Figure 39: Means and Standard Errors for perceived spanner length values in each
condition

61
Self-Efficacy Questionnaire

Cronbach’s alpha α, reliability measure was calculated for each condition to test
reliability of the self-efficacy scale of strength. The scale was highly reliable in
each condition (table 7). Means and Standard Errors were calculated for each
variable. The Shapiro-Wilk test showed normal distribution of the scale (all
p>.05). The ANOVA demonstrated no significant sound effects or interaction.
However, there was a significant combined effect of sound and repetition,
F(2,42)=3.69, p=.033. Thus, planned t-tests were performed separately for
repetition 1 and 2. A significant effect of sound was found in HF2
(M=42.56,SE=3.19) with respect to NF2 (M=39.68,SE=2.77), t(21)=-
2.20,p=.039,r=.43 and in HF2 with respect to LF2 (M=40.26,SE=3.01), t(21)=-
2.11,p=.047,r=.43.

Cronbach’s α
NF .868
NF2 .781
LF .823
LF2 .839
HF .847
HF2 .866
Table 7: Cronbach’s α for self-efficacy values in each experimental trial.

62
 Perceived Strenght - Sound Repetition 1
46

44

42
perc_strength__NF1
40 perc_strength__LF1
perc_strength__HF1
38

36

34

Figure 40: Means and Standard Errors for perceived strength values in sound repetition 1 -
for the measurement a scale from 0-100 was utilised, where 0 was “Cannot lift at all” and
100 was “Highly certain I can lift” a specific weight

Perceived Strength – Sound Repetition 2

50
45
40
35
30 perc_strenght__NF2
25 perc_strength__LF2
20 perc_strength__HF2
15
10
5
0

Figure 41: Means and Standard Errors for perceived strength values in sound repetition 2 -
for the measurement a scale from 0-100 was utilised, where 0 was “Cannot lift at all” and
100 was “Highly certain I can lift” a specific weight

63
4.2 Behavioral and Physiological Measures

Pressure sensors

Primarily due to technical implications only the data gathered from the sensors
integrated in the left shoe were analysed. More specifically, while the
experiment was in progress, the researcher observed that the right foot sensors
were giving significantly lower values than the left foot ones. However,
behavioural differences across conditions could still be noticed. Thus, it was
decided to run the Shapiro-Wilk test for normality and the ANOVA tests for
both feet data to confirm the validity of the initial assumption to exclude the
specific sensors. The results showed inconsistencies between the results of right
and left foot. For instance, while in certain cases significant sound effects were
revealed on the left foot data, no significant effects were found on the right foot.
Moreover, data were normal for most of the left foot variables but not for the
right foot. Finally, the fact that the accelerometer was attached to the left foot
also pointed to the direction of conducting the analysis in the sensors of the same
foot. Considering the aforementioned, it was concluded that the right foot
sensors should be excluded from the analysis.

There were seven measures of interest per step. Means were calculated for each
variable in each condition. The measures and their labeling are demonstrated in
Table 8.

A different number of steps were extracted and analysed between participants.

Six steps were analysed in seven participants. Five steps were analysed in seven
participants and four steps in one participant. Five participants took either five or
six steps and two participants either four or five steps across different conditions.
In the cases of having an uneven number of steps between conditions no steps
were excluded, since the number of steps could be related to the participants’
speed and important information could be lost. Additionally, due to technical
problems behavioural data were lost in NF1 condition for participant 7, NF2 for
participant 14, LF1 for participant 15 and HF1 for participant 19.

64
Overall, significant effects of sound were found in 2 out of 7 measures, average
time of heel contact with the ground and average of toe pressure applied on the
ground. No significant sound effects were found in the rest of the measures. For
the detailed analysis of the non-significant values please see Appendix G.

Measures - Heel Labelling

Average time of heel contact with the left_heel_t1t0_diff
ground
Average step time – time interval heel_to_toe
between heel’s initial strike and toe’s
last contact with the ground
Average of the maximum heel left_heel_peak
pressure
Average of heel pressure applied on left_heel_average
the ground
Measures - Toe
Average time of toe contact with the left_toe_t1t0_diff
ground
Average of maximum toe pressure left_toe_peak
Average toe pressure applied on the left_toe_average
ground
Table 8: Labelling of the behavioural measures

Condition Measure
NF1 NF1_left_heel_t1t0_diff
NF1_heel_to_toe
NF1_left_heel_peak
NF1_left_heel_average
NF1_left_toe_t1t0_diff
NF1_left_toe_peak
NF1_left_toe_average
NF2 NF2_left_heel_t1t0_diff
NF2_heel_to_toe

65
 NF2_left_heel_peak
NF2_left_heel_average
NF2_left_toe_t1t0_diff
NF2_left_toe_peak
NF2_left_toe_average
LF1 LF1_left_heel_t1t0_diff
LF1_heel_to_toe
LF1_left_heel_peak
LF1_left_heel_average
LF1_left_toe_t1t0_diff
LF1_left_toe_peak
LF1_left_toe_average
LF2 LF2_left_heel_t1t0_diff
LF2_heel_to_toe
LF2_left_heel_peak
LF2_left_heel_average
LF2_left_toe_t1t0_diff
LF2_left_toe_peak
LF2_left_toe_average
HF1 NF1_left_heel_t1t0_diff
NF1_heel_to_toe
NF1_left_heel_peak
NF1_left_heel_average
NF1_left_toe_t1t0_diff
NF1_left_toe_peak
NF1_left_toe_average
HF1 NF2_left_heel_t1t0_diff
NF2_heel_to_toe
NF2_left_heel_peak
NF2_left_heel_average
NF2_left_toe_t1t0_diff
NF2_left_toe_peak

66
 NF2_left_toe_average
Table 9: Variables in each condition

Average time of heel contact with the ground

The Shapiro-Wilk test showed normal data distribution of the scale. An ANOVA
was run for the left foot sensors. Mauchly’s test indicated that the assumption of
sphericity has been violated, (2)=9.67, p=.008. Therefore, degrees of freedom
were corrected using Greenhouse-Geisser estimates of sphericity, ε=.68. There
was a significant effect of sound, F(1.37,23.38)=3.95, p=.047. No repetition
effects or interaction were indicated (p>.05). A paired sample t-test revealed that
the duration of heel contact tended to be longer in LF (M=57.79,SE=1.57) than
in HF (M=55.96,SE=1.23) , t(21)=-2.07, p=.051<0.1, r=.41 and
NF(M=56.24,SE=1.28), t(21)=2.03, p=.055<0.1, r=.40. No significant effects
were found in HF with respect to NF condition, p>.05. The initial data were
further explored to detect outliers. However, no outlier data were identified.

Seconds (s) Time of heel contact

60
0.82
0.8
59
0.79
58
0.78
57
NF_heel_t1t0diff
0.76
56
LF_heel_t1t0diff
0.75
55 HF_heel_t1t0diff
54
0.73
53
0.72
52
0.71
NF_heel_t1t0diff LF_heel_t1t0diff HF_heel_t1t0diff

Figure 42: Means and Standard Errors for time of heel contact in each condition

Average of toe values

Shapiro-Wilk test showed non-normal data distribution, p<.05. Normalization

was achieved by converting the original data to z-scores. Thus, the z-scores were
used for subsequent analysis. No violation of assumption of sphericity was

67
found, p>.05 thus we are allowed to report the sphericity assumption,
F(2,34)=4.51, p=.018 which showed a significant effect of sound on average toe
values. A paired t-test revealed that the pressure applied on the floor by the ball
of the foot (toe area) was higher in LF (M=552.86,SE=15.88) than in HF
(M=543.48, SE=16.95), t(21)=-2.21, p=.038, r=.43. No significant sound effects
were found in NF with respect to LF and HF (all p>.05).

Average Heel Pressure on the ground

580
570
560
550 NF_left_toe_avg_mean
ohms (Ω) 540 LF_left_toe_avg_mean
530 HF_left_toe_avg_mean
520
510
500

Figure 43: Means and Standard Errors for the average of toe values in each condition – the
original non-transformed values have been used for the graphical representation- the
pressure in Force-Sensitive resistors is measured in ohms which can be converted to
measured Force (g).

Accelerometer

There were three measures of interest. Means were calculated for each variable
in each condition. The measures and their labeling are demonstrated in Table 10.
Due to technical implications data were lost in NF1 condition for participant 7,
in LF1 condition for participant 15 and in HF1 condition for participant 19.
Overall, the results suggested that there were significant sound effects in one
measure, the acceleration during lifting the foot to move forward. No significant
sound effects were found in the rest of the measures. For the analysis of the non-
significant measures please see Appendix G.

68
Measures
Acceleration during lifting the foot to acc_lifting
move forward
Acceleration when moving the foot acc_down
downwards towards the floor.
Deceleration occurred during lifting the dec_lifting
foot
Table 10: labelling of the behavioural measures

Sound Conditions Measures

NF1 NF1_acc_lifting
NF1_acc_down
NF1_dec_lifting
NF2 NF2_acc_lifting
NF2_acc_down
NF2_dec_lifting
LF1 LF1_acc_lifting
LF1_acc_down
LF1_dec_lifting
LF2 LF2_acc_lifting
LF2_acc_down
LF2_dec_lifting
HF1 HF1_acc_lifting
HF1_acc_down
HF1_dec_lifting
HF2 HF2_acc_lifting
HF2_acc_down
HF2_dec_lifting
Table 11: Variables in each condition

69
Acceleration when lifting the foot to move forward

Exploratory analysis revealed a normal data distribution. Mauchly’s test showed

no violation of assumption of sphericity. No significant effects of sound were
found F(2,36)=1.94, p=.158. The data were further explored to detect outliers by
looking for values that exceed or were under two SD from the mean. No outlier
data were detected. Since the p value was not far from significance and the
researcher’s initial hypothesis was that there will be a difference in the
acceleration values between HF and LF conditions, t-tests were carried out. The
results suggested that foot acceleration was significantly larger in HF
(M=2085.65,SE=49.60) than in LF (M=2048.41,SE=55.15), t(21)=2.30, p=.032,
r=.44. No significant sound effects were found in the rest paired comparisons
(all p>.05).

Accelation when lifting foot to move forwards

2150

2100

2050 NF_acc_lifting
g-force (g) LF_acc_lifting
2000 HF_acc_lifting

1950

1900

Figure 44: Means and Standard Errors for acceleration when lifting foot to move forward –
measurement unit

Galvanic Skin Response – GSR

Data from participant 2 were excluded from analysis due to noise in the GSR
signal. Thus, data from 21 instead of 22 participants were included in the
analysis. Additionally, data collected from participant 1 in HF2 and LF2
conditions were excluded for the same reason. The measures and their labeling
are demonstrated in Table 12. Overall, the results suggested that there were

70
significant sound effects in one measure, the difference between maximum and
minimum GSR during marching. No significant sound effects were found in the
rest of the measures. For the analysis of the non-significant measures please see
Appendix G.

A Shapiro-Wilk test of normality revealed that all the GSR measures were
significantly non-normally distributed, p<.05. None of the four transformations
mentioned in the beginning of the analysis chapter successfully normalized the
data distributions. However, since GSR is a highly variable measure not only
between participants, but also within participants, due to varying level of
moisture in the skin due to weather conditions, all the original data were
individually z-scored to control individual variability.

Measures Labelling
Average arousal during marching avg_march
Average arousal during walking avg_walk
Difference between maximum and max_min_march
minimum arousal during marching
Difference between maximum and max_min_walk
minimum arousal during walking
Table 12: labelling of the GSR behavioural measures

Sound Conditions Measures

NF1 NF1_avg_march_z
NF1_avg_walk_z
NF1_max_min_march_z
NF1_max_min_walk_z
NF2 NF2_avg_march_z
NF2_avg_walk_z
NF2_max_min_march_z
NF2_max_min_walk_z
LF1 LF1_avg_march_z
LF1_avg_walk_z
LF1_max_min_march_z

71
 LF1_max_min_walk_z
LF2 LF2_avg_march_z
LF2_avg_walk_z
LF2_max_min_march_z
LF2_max_min_walk_z
HF1 HF1_avg_march_z
HF1_avg_walk_z
HF1_max_min_march_z
HF1_max_min_walk_z
HF2 HF2_avg_march_z
HF2_avg_walk_z
HF2_max_min_march_z
HF2_max_min_walk_z
Table 13: Variables in each condition

Difference between maximum and minimum GSR during marching

The data distribution was non-normal. A Wilcoxon test was run between the two
repetitions of each condition to identify possible effects of sound repetition.
Since no significant effect was found the values of the 2 repetitions for each
condition were merged and their Means were calculated (Figure 45). A
Friedman’s test was run to detect significant effects of sound in the difference
between the minimum and maximum arousal. The results indicated a significant
sound effect, (2)= 9.57,p=.008. Paired comparisons between the three sound
conditions were carried out by performing a Wilcoxon test. The results indicated
that there was a significant effect of sound in the difference of the maximum and
minimum arousal values in LF (M = .18,SE = .09) with respect to HF (M =
.13,SE = .05), z=-2.13, p=.033, r=-.32 and in NF (M = .15,SE = .06) with respect
to HF conditions (M = .13,SE = .05), z=-2.09, p=.036, r=-.31. No significant
effects were identified in NF with respect to LF conditions, p>.05.

72
 Difference between maximum and minimum
arousal values
0.3

0.25
NF_max_min_march_av
0.2 erage
LF_max_min_march_ave
0.15
rage
0.1 HF_max_min_march_av
erage
0.05

Figure 45: Means and Standard errors of max – min difference in each condition –the
means of the original values are illustrated

73
 5 DISCUSSION, FUTURE WORK AND LIMITATIONS

5.1 Results Discussion

The hypotheses suggested in Chapter 2 are discussed in the light of the
experimental results. Please note that hypothesis 4 regarding walking posture
was not examined since the measures utilized for this purpose were not analysed
due to time limitations.

Hypothesis 1

Hypothesis 1
1. Participants’ perception of their own body weight will change with two
possible outcomes (Li et al.,1991):
1.1 Participants will perceive their own body weight as bigger when the low
frequency components of their footsteps sound increase.
1.2 Participants will perceive their own body weight as smaller when the
high frequency components of their footsteps sound increase.

To verify the effect of manipulated footstep sounds in perceived weight two

measures were analysed, the values of perceived weight (Body Visualization
Application) and the heavy-light 7-point likert scale.

Considering the Body Visualizing measure, participants perceived their body

weight as lower in HF condition than in NF condition. These findings extend Li
et al’s.,(1991) results which indicated that manipulated footstep sounds can
change the perception of someone else’s body weight, by demonstrating that
manipulated self-produced footstep sounds can change the perception of our own
body weight. In contrast with Li et al.,(1991), participants did not feel
significantly heavier in LF than in NF condition. This may be attributed to the
fact that the sound manipulations utilized deviated from the ones performed by
Li et al.,(1991). More specifically, while Li et al.,(1991) in their study increased
one low frequency band per time and decreased the rest ones, we proceeded to
the increment of multiple low frequency bands (63, 125 and 250 Hz). Our results

74
may indicate that the manipulations we performed were quite crude and the
sound output did not feel self-produced or participants were too aware of these
manipulations and they were not influenced. However, a trend revealed that
participants felt heavier in LF than in HF condition.

Looking at the second self-report measure the results show a degree of similarity
with the Body Visualization one. The findings suggest that participants felt
lighter in HF than in LF condition. A trend also showed that participants felt
lighter in HF condition than in NF condition. Similarly to the first measure there
were no changes in perceived body weight between the LF and NF condition
confirming our speculations about performing crude manipulations in LF.

Another interesting finding revealed from both measures is the fact that
participants felt heavier in the NF condition. This may indicate that our reference
condition (NF) was not appropriately chosen. While the acoustic feedback
provided during NF condition was not manipulated, it cannot be considered as
veridical since it is amplified and rendered through headphones. The results
indicate that the amplification alone affected perceived weight and was more
evident than the LF manipulation. This is not surprising since past studies
(Zampini and Spence,2005; Guest et al.,2002; Jousmaki and Hari,1998) have
demonstrated that overall sound amplification leads to changes in perceived
sound source characteristics.

Overall, the above mentioned results seem to confirm hypothesis 1. While

people receive multisensory feedback (tactile, visual, proprioceptive and
auditory) during walking our study showed that changing the auditory feedback
alone alters perceived body weight. These results complete past researches
which investigated the contribution of feedback of different modalities to body
representation (De Vignemont et al,2005; Haggard et al.,2008; Tajadura-Jimenez
et al’s., 2012) and highlight the significance of auditory feedback.

Hypothesis 2

Hypothesis 2
2. Participants’ motor behaviour will be affected (“Sound and Motor
Behaviour” section) by the provision of manipulated footsteps sounds:

75
 2.1 Participants will decrease their speed (longer time interval between heel
strike and toe-off events) when we increase the low frequency
components of the footsteps sound. It is also expected that both pressure
applied on the ground and time of contact with the ground will be
increased. Additionally, the acceleration while lifting the foot to move
forward will be decreased (point-light walker application, Troje,2008).
2.2 Participants will increase their speed (shorter time interval between heel
strike and toe-off events) when we increase the high frequency
components of the footsteps sound. It is also expected that both pressure
applied on the ground and time of contact with the ground will be
decreased. Additionally, the acceleration while lifting the foot to move
forward will be increased (point-light walker application, Troje,2008).

Participants’ motor behaviour was analysed in 4 dimensions, speed, pressure

applied on the ground, time of contact with the ground and acceleration of foot
movement.

Participants’ speed was measured by calculating the time interval between left
foot’s heel strike and toe-off events. The results showed that the acoustic
feedback did not significantly change participants’ speed. Thus this part of the
hypothesis is disconfirmed. These results are not consistent with previous
researches which demonstrated effects of auditory feedback in participants’
speed/pace (Menzer et al.,2010; Bresin et al.,2010). However possible changes
could have been revealed if we had the time to analyse more variables for
detecting changes in speed (e.g. time interval between 2 heel-strikes).

Regarding the pressure applied on the ground, while there are no sound effects
on the heel pressure, the findings reveal that the pressure applied from the ball of
the foot was higher in LF than in HF condition confirming this part of the
hypothesis. It is not surprising that the pressure changes were detected in the area
of the ball of the foot, since the relationship between these two factors is well
established in past research. More specifically, Zhu et al., (1991, as cited in

76
Titianova et al.,2004) in their study state that the heel-off event is followed by a
phase were the entire body weight is borne on the ball of the foot and toe area.
Since participants’ perceived auditory weight, as measured with the body
visualization application and the questionnaire, did not significantly changed in
the LF condition as described above, these results may indicate an alteration in
their unconscious “body-image” (De Vignemont, 2010) which led to increasing
the pressure exerted on the ground. On the other hand, changes in participants’
consciously perceived body weight are consistent with the applied pressure
regarding the HF condition. The fact that the pressure changes in HF and LF are
not significant with respect to NF condition underlines again our considerations
about the nature of the sound furnished during the reference condition (NF).

Regarding the time of contact with the ground no significant effects of sound
were found. However, there is a trend indicating that participants kept their heel
longer on the ground in LF than in HF condition. Another trend also indicates
that heel time contact was longer in LF than in NF. These results partially
confirm the Hypothesis.

Finally, concerning the acceleration of movement, while sound did not

significantly affect the deceleration and acceleration during the ascending and
descending foot movement respectively, significant effects were showed in the
acceleration during ascending the foot to move forward. More specifically, the
lifting acceleration was higher in HF than in LF conditions. Thus, our hypothesis
regarding the acceleration of foot is confirmed. These findings were expected
and consistent with our observations in Troje’s (2008) point-light walker
application.

Overall, Hypothesis 2 was partially confirmed since the pressure applied on the
ground, time of foot contact and acceleration of foot changed according to our
assumptions. However, no changes were detected in participants’ speed. These
results complete previous research suggesting that walking related sounds
(ground texture sounds) affect the walking pattern of people with specific
emotional intentions (Bresin et al.,2010) by demonstrating that auditorily
perceived weight cues in walking sounds can affect walking pattern. While we
manipulated body weight related auditory cues there is a possibility that people

77
perceived the sound output as changes in the ground or shoe materials, especially
when we performed high frequency manipulations. This assumption is based on
past studies which demonstrated that overall amplification or high frequency
manipulations alter the perceived sound source characteristics (Zampini and
Spence,2005; Guest et al.,2002; Jousmaki and Hari,1998).

Hypothesis 3

Hypothesis 3
3. Participants’ emotional experience will be affected by the provision of
the manipulated footsteps sounds with two possible outcomes (Tonetto et
al.,2014; Tajadura-Jimenez et al.;2008):
3.1 Participants will feel more negative/unaroused/submissive when we
increase the low frequency components of the footstep sounds.
3.2 Participants will feel more positive/aroused/dominant when we increase
the high frequency components of the footsteps sound.

Participants’ emotional reactions due to the provision of footstep feedback were

explored with the utilization of self-report (SAM) and physiological (GSR)
measures.

Looking at the SAM measure, footsteps sound significantly affected dominance

and valence and arousal. These results are consistent with past studies exploring
the effects of other types of auditory feedback in these emotional dimensions
(Tonneto et al.,2014; Serafin et al.,2011; Furfaro et al., 2013).

More specifically, participants felt happier and more positive in HF than in NF

condition. A trend also indicates that they have more positive emotions in HF
than in LF. A trend also revealed that participant felt more aroused in HF than in
NF. Since there are no sound effects in emotional arousal in LF condition, our
hypothesis is partially confirmed. In terms of perceived dominance participants
felt more dominant HF than in LF condition. A trend also suggests that they also
felt more in control in HF than in NF. These results confirm our hypothesis and
are consistent with Tonetto et al’s., (2014) study which revealed that participants

78
felt more dominant when presented with more acute sounds (elicited when
people walk in ceramic floors with leather shoes) than with quitter sounds
(elicited when walking in carpet with leather shoes). The fact that participants
felt less dominant in the LF condition may be attributed to the nature of the low
frequency feedback. Increasing the low frequency bands the sound produced was
heavy and resounding. This may have led to participants feeling unfit in a body
that produces this type of sounds.

Regarding the physiological measurement there was no effect of sound in skin

conductance during the walking part of the trials. However, a significant effect
was detected in the difference between the min and max GSR values of the
marching session. More specifically, there was a greater variation in GSR values
in LF than in HF condition and in HF than in NF conditions. These findings
suggest that participants’ emotional activation highly varied during the marching
session in the LF condition. However, this could possibly be a novelty effect
which faded away as the participants continued to the walking session.

Hypothesis 5

Hypothesis 5
5. Participants’ perception of their strength will be affected by the provision
of manipulated footstep sounds (Furfaro et al., 2013) with two possible
outcomes:
5.1 Participants will feel stronger as they perceive their body as heavier when
we increase the low frequency components of the footsteps sound.
Accordingly, they will feel weaker when we increase the high frequency
components.
5.2 Participants will feel weaker as they will feel unfit to a body that is
perceived as heavier when we increase the low frequency components of
the footsteps sound. Accordingly, they will feel stronger when we
increase the high frequency components.

To explore the effect of the provision of footstep sounds Bandura’s self-efficacy

scale and a weak-strong 7-point likert scale were employed.
79
The results of the self-efficacy scale support our initial hypothesis, since
participants perceived their strength as greater in the second repetition of
HF(HF2) than in the NF(NF2) and in HF(HF2) than in LF(LF2) condition.
Similar effects were also revealed in the likert scale. Participants reported that
they felt stronger in HF condition and weaker in LF condition. These findings
are also consistent with previous studies showing the effects of sound in
proprioception (Tajadura-Jimenez et al., 2012). The fact that participants felt
stronger in the HF condition may rather suggest that the HF sound was more
evident than the LF one or that in the LF condition participants may have felt
unfit in a body that was perceived as heavier and thus felt weaker.

Hypothesis 6

Hypothesis 6
6. Considering the relationship between body weight and walking posture
(point-light walker, Troje, 2008) we believe that participants’ perception
of their walking posture (straight-stooped/hunched) will be affected by
the provision of manipulated footstep sounds (Tajadura-Jimenez et al.,
2012) with two possible outcomes:
6.1 Participants will perceive their posture as stooped when the low
frequency components of the footsteps sounds increase.
6.2 Participants will perceive their posture as straight when the high
frequency components of the footsteps sounds increase.

Effects of sound in the perceived walking posture were explored with the
assessment of a stooped/hunched-straight 7-point likert scale. The results seem
to partially confirm our hypothesis. More specifically, participants perceived that
their walking posture was straighter in the second repetition of HF (HF2) than in
the second repetition of NF (NF2) condition. These findings agree with
Tajadura-Jimenez’s et al’s.,(2012) research regarding the effects of sound in
proprioception. Additionally, they are consistent with the observations we made
in Troje’s (2008) point-light walker application. However, the fact that the
effects were evident only in the second repetition of the sound conditions may

80
indicate that changes in perceived posture are evident after longer exposure to
sound feedback.

Hypothesis 7

Hypothesis 7

7. Based on the findings in “Sound and Motor Behaviour” section which

suggest that sound affects motor behaviour (speed) (Bresin et al., 2010)
and Menzer et al’s., (2010) study which reveals that sound affects
perceived motor behaviour we also believe that participants’ perceived
speed will be affected by the provision of manipulated footstep sounds
with two possible outcomes:
7.1 Participants’ perceived speed will be decreased when we increase the low
frequency components of the footsteps sound.
7.2 Participants’ perceived speed will be increased when we increase the high
frequency components of the footsteps sound.

For the exploration of participants’ perceived speed a slow-quick 7-point likert

scale was utilized. Interestingly, the results demonstrated that participants felt
quicker in HF condition than in NF and LF conditions. These findings confirm
our initial hypothesis and are in agreement with previous research results
(Menzer et al., 2010) which indicate that auditory cues can alter one’s perceived
motor behaviour. However, they are not in agreement with the results revealed
from the analysis of participants’ walking speed. While participants’ walking
behaviour in terms of speed did not change, alterations were revealed in
perceived speed. This findings may suggest that participants felt more fit in HF
condition.

Other explorative measures

The analysis of the measure introduced to assess participants’ ability to localize

their feet suggested that participants felt more confident to localize their feet in
HF and less confident in NF and LF conditions. These results could suggest that

81
during the NF and LF conditions participants’ proprioception was influenced by
the provision of footsteps sound in accord to previous research (Tajadura-
Jimenez et al., 2012). Additionally, we could also assume that since participants
felt more dominant/in control in HF condition they were more confident to
localize their body parts.

Participants’ ability to identify that they were producing the footstep sounds was
also assessed. The results showed no significant effects of sound. The mean
values indicate that participants mostly identified that the sounds were produced
by themselves. These results are not consistent with Menzer et al.,(2010) study.
While they found that delaying footsteps sound caused changes on participants’
feeling of being producing the sounds, our manipulations intending to evoke the
sensation that sounds are produced by a body of different weight did not affect
the agency for the participant’s body.

Finally, whether participants experienced surprising and unexpected or vivid

feelings of body was explored. There were no significant effects of the provision
of footsteps sounds. These results are not consistent with Tonetto’s et al.,(2014)
which showed that walking-related sound affects this sort of bodily sensations.

82
5.2 Applications in HCI and Design recommendations

Applications in HCI
In Chapter 2 the potential of the use of footsteps auditory feedback in various
fields (VR, games, rehabilitation and fitness) was discussed drawing on past
researches. In light of the above mentioned findings, possible applications of the
results will be presented. The study findings demonstrate that the provision of
manipulated footsteps sound can influence the perception of one’s own body
weight, motor behavior and emotions.

More specifically, our study suggests the provision of footsteps sound with
increased high frequencies results in people feeling lighter, quicker, stronger,
more aroused, happier and in control. Our findings also revealed that
acceleration during lifting the foot increases. This can be related to increased
performance. These effects along with the extensive use of wearable audio
devices and the integration of multiple sensors which record body activity into
smartphones suggest that a refined and compact version of our prototype can be
used for the development of fitness and rehabilitation applications. For instance,
a system which would be comprised of a pair of compact microphones, wireless
headphones and a compact equalizer could be connected to a smartphone
application and be used for walking, running or physical rehabilitation
applications that could increase the users’ performance, creating a more
pleasurable running experience and providing motivation.

Accordingly, the auditory feedback of footsteps sound that changes perceived

body weight and influences motor behavior and emotions could be utilized in
VR applications and games to enhance presence and enable a more realistic
interaction in various cases by making the user feel lighter or heavier, changing
motor behavior and triggering affective responses.

Design recommendations
The effects of each of the three different auditory feedbacks on auditory
perceived weight, motor behavior and emotions are summarized and presented in
the Table 14 below. They can be utilized as recommendations for the design of

83
systems/applications which aim to use the footsteps auditory feedback to change the user’s perception of own body weight, motor behavior or
trigger emotional responses.

Sound Own body Motor Emotions Perceived Perceive Perceived

Conditions weigh Behaviour Strength posture speed
perception
NF People will Users will Users will feel Users will Users will Users will feel
(Natural feel heavier keep their more unaroused feel weaker feel that slower when
footstep when feet longer on and more when they walk presented with
sound presented the ground negative when presented more NF than with
amplified) with NF when presented with with NF stooped HF footsteps
than with presented NF than with than with when sounds.
HF sounds. with NF than HF footsteps HF presented
with HF sound. footsteps with NF
footstep sound. than with
sound. HF sound.

HF (High Users will Users will Users will feel Users will Users will Users will feel
frequency feel lighter apply less more aroused, feel feel that quicker when
component when pressure and more pleased stronger they walk presented with

84
s increased presented keep their and more in when more HF than with
and low with HF feet shorter control when presented upright NF and LF
decreased) footstep on the ground presented with with HF when footsteps
sounds than when HF than with than with presented sounds.
with NF or presented NF footsteps NF and LF with HF
with LF with HF than sound. footsteps than with
footsteps with LF sound. NF
sounds. footsteps Pleasure and footsteps
sound. feeling of sound.
control will also
Foot be increased
acceleration when presented
will be higher with HF than
when with LF
presented footsteps sound.
with HF than
with LF
footstep
sounds

LF (Low Users will Users will Users will feel Users will Users will feel

85
frequency feel heavier apply more less pleasure and feel weaker slower when
component when pressure and less in control when presented with
s increased presented keep their when presented presented LF than with
and high with LF feet longer on with LF than with LF HF footsteps
decreased) footstep the ground with HF than with sounds.
sounds than when footsteps sound. HF
with HF presented footsteps
ones. with LF than sound.
with HF
footstep
sound.

User’s foot
acceleration
will be lower
when
presented
with LF than
with HF
footstep
sounds.
Table 14: Summary of finding

86
5.3 Limitations and Future Research
The main limitations of the present study concern the equipment and the
experimental setup. Our initial goal was to achieve the development of a portable
wireless system. While this was managed for most of the equipment, the
equalizer could not be converted to a wireless device due to technical
implications. As a result, a large cable exited the backpack to connect the device
into the power plug. The contact of the cable with the ground introduced
background noise. Additionally, this large cable along with the presence of
cables connecting the shoes to the Arduino board may have influenced
participants’ walking pattern, even if participants reported no discomfort when
they were asked. Furthermore, due to the weather conditions (high temperature)
during data collection most of the participants’ skin was sweaty. This caused
difficulties in the setup of the wireless EMG sensors/accelerometer. Overall, we
could say that the unnatural experimental setting may have introduced bias in the
data collected. Finally, time-limitations prevented us from conducting the
analysis on the data gathered from both lower limbs, EMG in lower limbs, and
body part inclinations (delsey sensors).

Future research should be conducted with a wireless improved version of the

prototype and avoiding the limitations described above. The replicability of the
effects of sound in all three directions (perception of body weight, motor
behaviour and emotions) should be tested. Additionally, further analysis should
be performed in data gathered from both lower limbs and the ones collected to
detect changes in walking posture. The present study did not reveal any sound
effects on walking speed. However, considering previous research findings on
the effects of sound on speed (e.g. Bresin et al., 2010; Menzer et al., 2010)
further analysis should be conducted (e.g. measure the time interval between 2
heel strikes of the right and left lower limb). Furthermore, the study results
indicated that the reference condition (NF) should be altered. For instance, future
research may consider having a reference condition where participants will listen
to their natural footsteps sounds with no amplification and no use of headphones.
Performing less crude manipulation when increasing the low frequency

87
components of the footsteps sound should also be investigated. Additionally,
open-ended questions could be introduced in the study to further investigate the
possible sensations evoked due to the provision of the manipulated footstep
sounds. Finally, a more refined and compact version of our prototype could be
used to further explore the effects of self-produced manipulated footsteps sound
in perceived weight, motor behaviour and emotions in the wild.

88
 6 Conclusion
The main aim of this study was to investigate whether the provision of
manipulated footstep sounds affects the perception of one’s body weight, motor
behaviour and emotions.

A system which was comprised of a pair of sensored shoes prototype, audio

equipment and sensors attached to the body was built to provide participants
with auditory feedback and capture behavioural and physiological changes
during walking.

The initial hypotheses formulated were either confirmed or partially supported

by the study’s results. The research’s findings revealed effects of footstep sounds
in all three dimensions, auditory perceived weight of one’s body, motor
behaviour and emotions. More specifically, participants felt lighter when the
high frequency components of their footsteps sound were increased (HF sound).
Additionally, they felt heavier when they were presented with footsteps auditory
feedback in which the low frequency components were amplified (LF sound),
compared to the HF sound.

Furthermore, in terms of behavioural change, participants applied more pressure

on the ground and kept their foot longer on the ground when they were presented
with the LF sound than with the HF one. They also increased the acceleration
during ascending the foot to move forward when provided with the HF feedback
than with the LF feedback during which their foot acceleration was decreased.

Finally, considering changes in affective responses, participants reported that

they felt happier, more positive and more in control/dominant in HF than in LF
condition during which they felt more submissive and negative. Additionally,
participants felt more aroused in HF compared to NF condition.

In light of these findings design recommendations were formulated for the

integration of footstep sounds in interactive sound-based technologies. These
technologies have a wide applicability in domains such as fitness, rehabilitation,
VR and games. Future work includes examination of the results replicability and
the development of a more refined and compact prototype version to perform in
the wild studies.

89
REFERENCES

1. Ballas, J. A. (1993). Common factors in the identification of an

assortment of brief everyday sounds. Journal of experimental
psychology: human perception and performance, 19(2), 250.
2. Bandura, A. (1977). Self-efficacy: toward a unifying theory of behavioral
change.Psychological review, 84(2), 191.
3. Bandura, A. (2006). Guide for constructing self-efficacy scales. Self-
efficacy beliefs of adolescents, 5(307-337).
4. Becker, N., Brett, S., Chambliss, C., GROWERS, K., Haring, P., Marsh,
C., & Montemayor, R. (1994). Mellow and frenetic antecedent music
during athletic performance of children, adults, and seniors. Perceptual
and motor skills, 79(2), 1043-1046.
5. Botvinick, M., & Cohen, J. (1998). Rubber hands' feel'touch that eyes
see.Nature, 391(6669), 756-756.
6. Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: the self-
assessment manikin and the semantic differential. Journal of behavior
therapy and experimental psychiatry, 25(1), 49-59.
7. Bradley, M. M., & Lang, P. J. (2000). Affective reactions to acoustic
stimuli.Psychophysiology, 37(02), 204-215.
8. Bresin, R., de Witt, A., Papetti, S., Civolani, M., & Fontana, F. (2010).
Expressive sonification of footstep sounds. In Proceedings of ISon 2010:
3rd Interactive Sonification Workshop. Stockholm, Sweden: KTH Royal
Institute of Technology.
9. Cunado, D., Nixon, M. S., & Carter, J. N. (2003). Automatic extraction
and description of human gait models for recognition
purposes. Computer Vision and Image Understanding, 90(1), 1-41.
10. De Vignemont, F. (2010). Body schema and body image—Pros and
cons.Neuropsychologia, 48(3), 669-680.
11. De Vignemont, F., Ehrsson, H. H., & Haggard, P. (2005). Bodily
illusions modulate tactile perception. Current Biology, 15(14), 1286-
1290.
12. Ekimov, A., & Sabatier, J. M. (2006). Vibration and sound signatures of
human footsteps in buildingsa). The Journal of the Acoustical Society of
America,120(2), 762-768.
13. Fairburn, C. G. and Cooper, Z. (1993). The eating disorder
examination(12th edition). In: Fairburn, C. G. and G.T. Wilson (eds.).
Binge Eating: Nature, Assessment and Treatment. (pp 317-360). New
York: Guilford Press.
14. Ferguson, A. R., Carbonneau, M. R., & Chambliss, C. (1994). Effects of
positive and negative music on performance of a karate drill. Perceptual
and motor skills, 78(3c), 1217-1218.

90
15. Field, A. P. (2009). Discovering statists using SPSS (and sex and drugs
and rock ‘n’roll).
16. Freed, D. J. (1990). Auditory correlates of perceived mallet hardness for
a set of recorded percussive sound events. The Journal of the Acoustical
Society of America, 87(1), 311-322.
17. Furfaro, E., Bevilacqua, F., & Tajadura Jimenez, A. (2013). Sonification
of surface tapping: Influences on behaviour, emotion and surface
perception. Interactive Sonification Workshop.
18. Gaver, W. W. (1993). What in the world do we hear?: An ecological
approach to auditory event perception. Ecological psychology, 5(1), 1-29.
19. Giordano, B. L., Visell, Y., Yao, H. Y., Hayward, V., Cooperstock, J. R.,
& McAdams, S. (2012). Identification of walked-upon materials in
auditory, kinesthetic, haptic, and audio-haptic conditionsa). The Journal
of the Acoustical Society of America, 131(5), 4002-4012.
20. Giordano, B., & Bresin, R. (2006). Walking and playing: What’s the
origin of emotional expressiveness in music. In Proc. Int. Conf. Music
Perception and Cognition.
21. Guest, S., Catmur, C., Lloyd, D., & Spence, C. (2002). Audiotactile
interactions in roughness perception. Experimental Brain
Research, 146(2), 161-171.
22. Haggard, P., Christakou, A., & Serino, A. (2007). Viewing the body
modulates tactile receptive fields. Experimental Brain Research, 180(1),
187-193.
23. http://bodyvisualizer.com/
24. Jaffe, D. L., Brown, D. A., Pierson-Carey, C. D., Buckley, E. L., & Lew,
H. L. (2004). Stepping over obstacles to improve walking in individuals
with poststroke hemiplegia. Journal of rehabilitation research and
development,41(3), 283-292.
25. Jenkins, J. J. (1985). Acoustic information for objects, places, and
events.Persistence and change, 115-138.
26. Jousmäki, V., & Hari, R. (1998). Parchment-skin illusion: sound-biased
touch.Current Biology, 8(6), R190-R191.
27. Kitagawa, N., & Spence, C. (2006). Audiotactile multisensory
interactions in human information processing. Japanese Psychological
Research, 48(3), 158-173.
28. Lang, P. J. (1995). The emotion probe: Studies of motivation and
attention.American psychologist, 50(5), 372.
29. Lederman, S. J. (1979). Auditory texture perception. Perception, 8(1),
93-103.
30. Lederman, S. J., Klatzky, R. L., Morgan, T., & Hamilton, C. (2002).
Integrating multimodal information about surface texture via a probe:
relative contributions of haptic and touch-produced sound sources.
In Haptic Interfaces for Virtual Environment and Teleoperator Systems,

91
 2002. HAPTICS 2002. Proceedings. 10th Symposium on (pp. 97-104).
IEEE.
31. Leman, M., Moelants, D., Varewyck, M., Styns, F., van Noorden, L., &
Martens, J. P. (2013). Activating and relaxing music entrains the speed of
beat synchronized walking. PloS one, 8(7), e67932.
32. Li, X., Logan, R. J., & Pastore, R. E. (1991). Perception of acoustic
source characteristics: Walking sounds. The Journal of the Acoustical
Society of America, 90(6), 3036-3049.
33. Lunn, D., & Harper, S. (2010). Using galvanic skin response measures to
identify areas of frustration for older web 2.0 users. In Proceedings of the
2010 International Cross Disciplinary Conference on Web Accessibility
(W4A) (p. 34). ACM.
34. Menzer, F., Brooks, A., Halje, P., Faller, C., Vetterli, M., & Blanke, O.
(2010). Feeling in control of your footsteps: Conscious gait monitoring
and the auditory consequences of footsteps. Cognitive neuroscience, 1(3),
184-192.
35. Moens, B., van Noorden, L., & Leman, M. (2010). D-jogger: Syncing
music with walking. In 7th Sound and Music Computing Conference (pp.
451-456). Universidad Pompeu Fabra.
36. Montagu, J. D., & Coles, E. M. (1966). Mechanism and measurement of
the galvanic skin response. Psychological Bulletin, 65(5), 261.
37. Nordahl, R. (2006). Increasing the motion of users in photo-realistic
virtual environments by utilising auditory rendering of the environment
and ego-motion. In PRESENCE 2006: The 8th Annual International
Workshop on Presence (pp. 57-63).
38. Norman, D. A. (2002). The design of everyday things. Basic books.
39. Pastore, R. E., Flint, J. D., Gaston, J. R., & Solomon, M. J. (2008).
Auditory event perception: The source—perception loop for posture in
human gait. Perception & psychophysics, 70(1), 13-29.
40. Perala, C. H., & Sterling, B. S. (2007). Galvanic skin response as a
measure of soldier stress (No. ARL-TR-4114). ARMY RESEARCH
LAB ABERDEEN PROVING GROUND MD HUMAN RESEARCH
AND ENGINEERING DIRECTORATE.
41. Perry, J. (1992). Gait analysis. Normal and pathological function, 1.
42. Point-light walker application:
http://biomotionlab.ca/Demos/BMLwalker.html
43. Repp, B. H. (1987). The sound of two hands clapping: An exploratory
study.The Journal of the Acoustical Society of America, 81(4), 1100-
1109.
44. Rosati, G., Oscari, F., Reinkensmeyer, D. J., Secoli, R., Avanzini, F.,
Spagnol, S., & Masiero, S. (2011, June). Improving robotics for
neurorehabilitation: enhancing engagement, performance, and learning
with auditory feedback. InRehabilitation Robotics (ICORR), 2011 IEEE
International Conference on (pp. 1-6). IEEE.

92
45. Senna, I., Maravita, A., Bolognini, N., & Parise, C. V. (2014). The
Marble-Hand Illusion. PloS one, 9(3), e91688.
46. Serafin, S., Franinovic, K., Hermann, T., Lemaitre, G., Rinott, M.,
Rocchesso, D. (2011). Sonic interaction design. In T. Hermann, A. Hunt,
J. G. Neuhoff (Eds.), The sonification handbook (pp. 87-111). Berlin:
Logos Publishing House.
47. Singh, A., Klapper, A., Jia, J., Fidalgo, A., Tajadura-Jiménez, A.,
Kanakam, N., ... & Williams, A. (2014, April). Motivating people with
chronic pain to do physical activity: opportunities for technology design.
In Proceedings of the 32nd annual ACM conference on Human factors in
computing systems (pp. 2803-2812). ACM.
48. Styns, F., van Noorden, L., Moelants, D., & Leman, M. (2007). Walking
on music. Human movement science, 26(5), 769-785.
49. Suzuki, Y., Gyoba, J., & Sakamoto, S. (2008). Selective effects of
auditory stimuli on tactile roughness perception. Brain research, 1242,
87-94.
50. Tajadura-Jiménez, A., Väljamäe, A., & Västfjäll, D. (2008). Self-
representation in mediated environments: the experience of emotions
modulated by auditory-vibrotactile heartbeat. CyberPsychology &
Behavior, 11(1), 33-38.
51. Tajadura-Jiménez, A., Väljamäe, A., Toshima, I., Kimura, T., Tsakiris,
M., & Kitagawa, N. (2012). Action sounds recalibrate perceived tactile
distance.Current Biology, 22(13), R516-R517.
52. Titianova, E. B., Mateev, P. S., & Tarkka, I. M. (2004). Footprint
analysis of gait using a pressure sensor system. Journal of
Electromyography and Kinesiology,14(2), 275-281.
53. Tonetto, P. L. M., Klanovicz, C. P., & Spence, P. C. (2014). Modifying
action sounds influences people’s emotional responses and bodily
sensations. i-Perception, 5(3), 153-163.
54. Troje, N. F. (2008). Retrieving information from human movement
patterns.Understanding events: How humans see, represent, and act on
events, 1, 308-334.
55. Turchet, L., Nordahl, R., Serafin, S., Berrezag, A., Dimitrov, S., &
Hayward, V. (2010). Audio-haptic physically-based simulation of
walking on different grounds. In Multimedia Signal Processing (MMSP),
2010 IEEE International Workshop on (pp. 269-273). IEEE.
56. Väljamäe, A., Larsson, P., Västfjäll, D., & Kleiner, M. (2008). Sound
representing self-motion in virtual environments enhances linear
vection.Presence, 17(1), 43-56.
57. Vaughan, C. L., Davis, B. L., & O'connor, J. C. (1992). Dynamics of
human gait(pp. 15-43). Champaign, Illinois: Human Kinetics Publishers.
58. Visell, Y., Fontana, F., Giordano, B. L., Nordahl, R., Serafin, S., &
Bresin, R. (2009). Sound design and perception in walking

93
 interactions. International Journal of Human-Computer Studies, 67(11),
947-959.
59. Vlaeyen, J. W., & Linton, S. J. (2000). Fear-avoidance and its
consequences in chronic musculoskeletal pain: a state of the
art. Pain, 85(3), 317-332.
60. Warren, W. H., & Verbrugge, R. R. (1984). Auditory perception of
breaking and bouncing events: a case study in ecological
acoustics. Journal of Experimental Psychology: Human perception and
performance, 10(5), 704.
61. Zampini, M., & Spence, C. (2004). The role of auditory cues in
modulating the perceived crispness and staleness of potato chips. Journal
of sensory studies,19(5), 347-363.
62. Zanotto, D., Rosati, G., Spagnol, S., Stegall, P., & Agrawal, S. K. (2013).
Effects of complementary auditory feedback in robot-assisted lower
extremity motor adaptation. Neural Systems and Rehabilitation
Engineering, IEEE Transactions on, 21(5), 775-786.

94
APPENDICES

APPENDIX A - ARDUINO SCRIPT

/*

Code based on Tom Igoe's Serial Graphing Sketch

>> http://processing.org/exhibition/features/igoe/

Reads analog inputs and sends them as a series of strings seperated with commas,

and carriage returns after each full set of sensor values.

>> http://www.kobakant.at/DIY/?cat=347

*/

/*

MMA8452Q Basic Example Code

Nathan Seidle

SparkFun Electronics

November 5, 2012

License: This code is public domain but you buy me a beer if you use this and we meet someday
(Beerware license).

This example code shows how to read the X/Y/Z accelerations and basic functions of the
MMA5842. It leaves out

all the neat features this IC is capable of (tap, orientation, and inerrupts) and just displays X/Y/Z.
See

the advanced example code to see more features.

Hardware setup:

MMA8452 Breakout ------------ Arduino

3.3V --------------------- 3.3V

SDA -------^^(330)^^------- A4

SCL -------^^(330)^^------- A5

GND ---------------------- GND

The MMA8452 is 3.3V so we recommend using 330 or 1k resistors between a 5V Arduino and the
MMA8452 breakout.

95
The MMA8452 has built in pull-up resistors for I2C so you do not need additional pull-ups.

*/

#include <Wire.h> // Used for I2C

// The SparkFun breakout board defaults to 1, set to 0 if SA0 jumper on the bottom of the board is
set

#define MMA8452_ADDRESS 0x1D // 0x1D if SA0 is high, 0x1C if low

//Define a few of the registers that we will be accessing on the MMA8452

#define OUT_X_MSB 0x01

#define XYZ_DATA_CFG 0x0E

#define WHO_AM_I 0x0D

#define CTRL_REG1 0x2A

#define GSCALE 2 // Sets full-scale range to +/-2, 4, or 8g. Used to calc real g values.

int

numberOfSensors = 4;

int sensorPins [6] = {

0, 1, 2, 3, 4, 5}; //Analog input pins 0-5

int pullupPins [6] = {

14, 15, 16, 17, 18, 19}; //Digital names for analog input pins 0-5

void setup() {

for(int i = 0; i < numberOfSensors; i++) // set analog pins as inputs, although this is also the default

pinMode(sensorPins[i], INPUT);

for(int i = 0; i < numberOfSensors; i++) // set internal pullup resistors for all analog pins in use

digitalWrite(pullupPins[i], HIGH);

Serial.begin(57600);

Wire.begin(); //Join the bus as a master

initMMA8452(); //Test and intialize the MMA8452

96
void loop() {

int accelCount[3]; // Stores the 12-bit signed value

readAccelData(accelCount); // Read the x/y/z adc values

// Now we'll calculate the accleration value into actual g's

float accelG[3]; // Stores the real accel value in g's

for (int i = 0 ; i < 3 ; i++)

accelG[i] = (float) accelCount[i] / ((1<<12)/(2*GSCALE)); // get actual g value, this depends on

scale being set

for (int count = 0; count < numberOfSensors; count++) {

int sensorReading = analogRead(sensorPins[count]); // read each sensor as deffined in

sensorPins array

Serial.print(sensorReading, DEC); // print its value out as an ASCII numeric string

if (count < numberOfSensors - 1) Serial.print(","); // if this isn't the last sensor to read then print
a comma after it

// Print out values

for (int i = 0 ; i < 3 ; i++)

Serial.print(","); // commas in between axes

Serial.print(accelG[i]*1000, 2); // Print g values

Serial.println(); // after all the sensors have been read print a newline and carriage return

delay(10); // delay by # milliseconds

void readAccelData(int *destination)

97
 byte rawData[6]; // x/y/z accel register data stored here

readRegisters(OUT_X_MSB, 6, rawData); // Read the six raw data registers into data array

// Loop to calculate 12-bit ADC and g value for each axis

for(int i = 0; i < 3 ; i++)

int gCount = (rawData[i*2] << 8) | rawData[(i*2)+1]; //Combine the two 8 bit registers into one
12-bit number

gCount >>= 4; //The registers are left align, here we right align the 12-bit integer

// If the number is negative, we have to make it so manually (no 12-bit data type)

if (rawData[i*2] > 0x7F)

gCount = ~gCount + 1;

gCount *= -1; // Transform into negative 2's complement #

destination[i] = gCount; //Record this gCount into the 3 int array

// Initialize the MMA8452 registers

// See the many application notes for more info on setting all of these registers:

// http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=MMA8452Q

void initMMA8452()

byte c = readRegister(WHO_AM_I); // Read WHO_AM_I register

if (c == 0x2A) // WHO_AM_I should always be 0x2A

Serial.println("MMA8452Q is online...");

else

98
 Serial.print("Could not connect to MMA8452Q: 0x");

Serial.println(c, HEX);

while(1) ; // Loop forever if communication doesn't happen

MMA8452Standby(); // Must be in standby to change registers

// Set up the full scale range to 2, 4, or 8g.

byte fsr = GSCALE;

if(fsr > 8) fsr = 8; //Easy error check

fsr >>= 2; // Neat trick, see page 22. 00 = 2G, 01 = 4A, 10 = 8G

writeRegister(XYZ_DATA_CFG, fsr);

//The default data rate is 800Hz and we don't modify it in this example code

MMA8452Active(); // Set to active to start reading

// Sets the MMA8452 to standby mode. It must be in standby to change most register settings

void MMA8452Standby()

byte c = readRegister(CTRL_REG1);

writeRegister(CTRL_REG1, c & ~(0x01)); //Clear the active bit to go into standby

// Sets the MMA8452 to active mode. Needs to be in this mode to output data

void MMA8452Active()

byte c = readRegister(CTRL_REG1);

writeRegister(CTRL_REG1, c | 0x01); //Set the active bit to begin detection

// Read bytesToRead sequentially, starting at addressToRead into the dest byte array

void readRegisters(byte addressToRead, int bytesToRead, byte * dest)

99
 Wire.beginTransmission(MMA8452_ADDRESS);

Wire.write(addressToRead);

Wire.endTransmission(false); //endTransmission but keep the connection active

Wire.requestFrom(MMA8452_ADDRESS, bytesToRead); //Ask for bytes, once done, bus is

released by default

while(Wire.available() < bytesToRead); //Hang out until we get the # of bytes we expect

for(int x = 0 ; x < bytesToRead ; x++)

dest[x] = Wire.read();

// Read a single byte from addressToRead and return it as a byte

byte readRegister(byte addressToRead)

Wire.beginTransmission(MMA8452_ADDRESS);

Wire.write(addressToRead);

Wire.endTransmission(false); //endTransmission but keep the connection active

Wire.requestFrom(MMA8452_ADDRESS, 1); //Ask for 1 byte, once done, bus is released by

default

while(!Wire.available()) ; //Wait for the data to come back

return Wire.read(); //Return this one byte

// Writes a single byte (dataToWrite) into addressToWrite

void writeRegister(byte addressToWrite, byte dataToWrite)

Wire.beginTransmission(MMA8452_ADDRESS);

Wire.write(addressToWrite);

Wire.write(dataToWrite);

Wire.endTransmission(); //Stop transmitting

100
APPENDIX B - ARDUINO SCRIPT
/*

Serial Graphing Sketch

by Tom Igoe

Language: Processing

This sketch takes ASCII values from the serial port

at 9600 bps and graphs them.

The values should be comma-delimited, with a newline

at the end of every set of values.

The expected range of the values is between 0 and 1023.

Created 20 April 2005

Updated 27 June 2008

*/

import processing.serial.*;

PrintWriter output;

int maxNumberOfSensors = 7; // Arduino has 6 analog inputs, so I chose 6

boolean fontInitialized = false; // whether the font's been initialized

Serial myPort; // The serial port

float[] previousValue = new float[maxNumberOfSensors]; // array of previous values

int xpos = 0; // x position of the graph

PFont myFont; // font for writing text to the window

color[] palette; // colour palette

int lineThickness = 2;

//Variable to store the start time

int startTime = 0;

//Variable for saving milliseconds passed

int newTimer = 0;

//Variable for enable file recording

101
boolean startRecording = false;

//Time of recording

final int RECORDING_TIME = 60;

void setup () {

output = createWriter("Maria.txt");

// setup color palette

palette=new color[10];

palette[0]=color(250,0,0);

palette[1]=color(138,155,15);

palette[2]=color(48,139,206);

palette[3]=color(252,252,38);

palette[4]=color(255,255,255);

palette[5]=color(232,16,175);

palette[6]=color(247,134,12);

palette[7]=color(255,255,255);

palette[8]=color(255,255,255);

palette[9]=color(255,255,255);

// set up the window to whatever size you want:

size(1024, 697);

// List all the available serial ports:

println(Serial.list());

// I know that the first port in the serial list on my mac

// is always my Arduino or Wiring module, so I open Serial.list()[0].

// Open whatever port is the one you're using.

102
 String portName = Serial.list()[2];

myPort = new Serial(this, portName, 57600);

myPort.clear();

// don't generate a serialEvent() until you get a newline (\n) byte:

myPort.bufferUntil('\n');

// create a font with the fourth font available to the system:

//String[] fontList = PFont.list();

//println(fontList);

myFont = createFont(PFont.list()[3], 18);

textFont(myFont);

fontInitialized = true;

// set inital background:

background(0);

// turn on antialiasing:

smooth();

void draw () {

// nothing happens in the draw loop,

// but it's needed to keep the program running

void serialEvent (Serial myPort) {

// get the ASCII string:

String inString = new String(myPort.readBytesUntil('\n'));

// if it's not empty:

if (inString != null) {

// trim off any whitespace:

103
inString = trim(inString);

//println(inString);

// convert to an array of ints:

float incomingValues[] = float (split(inString, ","));

//output.println("incoming values: " + incomingValues.length);

// print out the values

// print("length: " + incomingValues.length + " values.\t");

if(startRecording){

output.println(inString);

if (incomingValues.length <= maxNumberOfSensors && incomingValues.length > 0) {

for (int i = 0; i < incomingValues.length; i++) {

// map the incoming values (0 to 1023) to an appropriate

// graphing range (0 to window height/number of values):

//float ypos = map(incomingValues[i], 0, 1023, 0, height/incomingValues.length);

float ypos = map(incomingValues[i], -2000, 2000, 0, height/incomingValues.length);

// figure out the y position for this particular graph:

float graphBottom = i * height/incomingValues.length;

ypos = ypos + graphBottom;

// make a black block to erase the previous text:

noStroke();

fill(0);

rect(10, graphBottom+1, 130, 24);

104
 // print the sensor numbers to the screen:

fill(palette[i]);

int textPos = int(graphBottom) + 22;

// sometimes serialEvent() can happen before setup() is done.

// so you need to make sure the font is initialized before

// you text():

if (fontInitialized) {

text("Sensor " + i + ":" + incomingValues[i], 10, textPos);

// draw a line at the bottom of each graph:

stroke(127);

line(0, graphBottom, width, graphBottom);

// change colors to draw the graph line:

//stroke(64*i, 32*i, 255);

stroke(palette[i]);

strokeWeight(lineThickness);

line(xpos, previousValue[i], xpos+1, ypos);

strokeWeight(1);

// save the current value to be the next time's previous value:

previousValue[i] = ypos;

// if you've drawn to the edge of the window, start at the beginning again:

if (xpos >= width) {

xpos = 0;

background(0);

else {

105
 xpos++;

// Ends the program after 1 minute

if(startRecording){

//Gets time since we pressed the key

newTimer = millis() - startTime;

println("time: " + newTimer);

//executes the exit function when the timer gets to the end

if((newTimer/1000) >= RECORDING_TIME){

goExit();

void goExit() {

output.println("Date End: " + day() + " " + hour() + ":" + minute() + ":" + second());

output.flush(); // Writes the remaining data to the file

output.close(); // Finishes the file

exit(); // Stops the program

void keyPressed(){

//Print the date in the file

output.println("Date: " + day() + " " + hour() + ":" + minute() + ":" + second());

// Tells everyone that they should start recording

startRecording = true;

// millis tells you how much time the program has been running so we need to substract it when
we start the program

startTime = millis();

106
APPENDIX C - QUESTIONNAIRES

Participant No:

Questionnaire 1

1. What is your gender?

Male

Female

2. What is your age?

18 – 24

25 – 34

35 – 44

45 – 54

55 – 64

65 years or older

3. What is your weight at present? (Please give your best estimate and circle the
unit of measurement).

………………. pounds / Kg

4. What is your height? (Please give your best estimate and circle the unit of
measurement).

………………. inches / cm

5. Do you have any music education?

107
 Yes

No

6. If yes please mention.

………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………

7. Do you exercise?

Yes

No

8. If yes how many hours per week do you exercise? (e.g. 5 hours)

……………………………...........................................................................................
................

9. Do you have any hobbies?

Yes

No

10. If yes please mention them.

………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………………………………………………
………………………………………………

108
 Condition:

Participant No:

Spanner

Now please think about the experience you have just had. You will have to answer a few
questions about it.

During the experience you were asked to carry a spanner:

1. What was the approximate weight of the spanner you carried? (Please give
your best estimate and circle the unit of measurement).

………………………….. ounces/ grams

2. What was the approximate length of the spanner you carried? (Please give
your best estimate and circle the unit of measurement).

………………………….. inches/ cm

3. Which hand did you use to carry the spanner?

Right

Left

109
 Condition:

Participant No:

Lifting

If you were asked to lift objects of different weights RIGHT NOW, how certain are you
that you can lift each of the weights described below?

Rate you degree of confidence by recording a number from 0 to 100 using the scale given
below. You can write any number from 0 to 100 (e.g. 42).

0 10 20 30 40 50 60 70 80 90 100

Cannot Moderately Highly certain

do at all can do can do

Physical Strength Confidence

(0-100)
Lift a 10 Pound (4.5 Kg) object
Lift a 20 Pound (9 Kg) object
Lift a 50 Pound (23 Kg) object
Lift a 80 Pound (36 Kg)object
Lift a 100 Pound (45 Kg) object
Lift a 150 Pound (68 Kg) object
Lift a 200 Pound (91 Kg) object
Lift a 300 Pound (136 Kg) object

110
 Condition:

Participant No:
Questionnaire 2.1

Now please circle the number you think that best expresses your level of agreement with
the sentences below.

During the experience I felt:

1 2 3 4 5 6 7
Slow Quick

During the experience I felt:

1 2 3 4 5 6 7
Light Heavy

During the experience I felt:

1 2 3 4 5 6 7
Weak Strong

During the experience it felt I was walking:

1 2 3 4 5 6 7
Stooped, Straight
Hunched

111
During the experience I felt the sounds I heard were produced by my own footsteps/
body.

1 2 3 4 5 6 7
I strongly Neither agree I strongly
disagree nor disagree agree

During the experience the feeling of my body was less vivid than normal.

1 2 3 4 5 6 7
I strongly Neither agree I strongly
disagree nor disagree agree

During the experience the feelings about my body were surprising and unexpected.

1 2 3 4 5 6 7
I strongly Neither agree I strongly
disagree nor disagree agree

During the experience I could really tell where my feet were.

1 2 3 4 5 6 7
I strongly Neither agree I strongly
disagree nor disagree agree

112
 Condition:

Participant No:
Questionnaire 2.2

How did you feel during the experience?

You will have to rate the way you felt by selecting a figure from each of 3 different scales.
Each scale shows different kind of feelings: happy/positive vs unhappy/negative,
aroused/excited vs unaroused/calm, dominant/important vs submissive/awed (slightly
frightened).

Unhappy, Negative Neutral Happy, Positive

Unaroused, Calm Neutral Aroused, Excited

Submissive, Awed Neutral Dominant, Important

113
 Participant No:

Questionnaire 3

The following questions are concerned with the past four weeks (28 days) only. Please read
each question carefully. Please answer all the questions. Thank you.

Please circle the appropriate number on the right. Remember that the questions only refer to
the past four weeks (28 days).

Over the past 28 Not at Slightly Moderately Markedly

days………. all
Have you had a
definite desire to
have a totally flat 0 1 2 3 4 5 6
stomach?
Has thinking
about shape or
weight made it
very difficult to
concentrate on 0 1 2 3 4 5 6
things you are
interested in (for
example
working,
following a
conversation or
reading)?
Have you had a
definite fear that
you might gain 0 1 2 3 4 5 6
weight?

Have you felt 0 1 2 3 4 5 6

fat?
Have you had a
strong desire to 0 1 2 3 4 5 6
lose weight?
Has your weight
influenced how
you think about 0 1 2 3 4 5 6
(judge) yourself
as a person?
Has your shape

114
influenced how
you think about 0 1 2 3 4 5 6
(judge) yourself
as a person?
How much would
it have upset you
if you had been
asked to weigh 0 1 2 3 4 5 6
yourself once a
week (no more,
or less often) for
the next four
weeks?
How dissatisfied
have you been 0 1 2 3 4 5 6
with your
weight?
How dissatisfied
have you been 0 1 2 3 4 5 6
with your shape?
How
uncomfortable
have you felt
seeing your body
(for example 0 1 2 3 4 5 6
seeing your shape
in the mirror, in
a shop window
reflection, while
undressing or
taking a bath or
shower)?

How
uncomfortable
have you felt
about others
seeing your shape
or figure (for 0 1 2 3 4 5 6
example in
communal
changing rooms,
when swimming
or wearing tight
clothes)?

115
APPENDIX D – CONCENT FORM

Informed Interviewee Consent Form

1. I hereby agree to participate in an experimental study in connection with research being conducted by Maria
Basia (MSc student at University College London), in collaboration with Dr Merle Fairhurst and Dr Ophelia
Deroy (Rethinking the Senses, CenSes).
2. Behavioural responses will be recorded in digital format. Additionally, relevant biographical information and
questionnaires in paper format will be stored in a restricted access, locked cabinet and used to analyse the data.
All data will be coded and identified by this code.
3. I understand that the testing will take no longer than 60 minutes and that I am free to withdraw at any stage. In
the event that I withdraw from the study, any data acquired will be destroyed.
4. I understand that, upon completion of the study, the data acquired may be referred to in research papers, reports
or conference proceedings albeit anonymously.

5. If I have questions about the research project or procedures, I know I can contact Dr Ana Tajadura-Jimenez,
supervisor and research fellow at the University College London Interaction Centre; email:
[email protected]

6. This study has been approved by the UCL Research Ethics Committee as Project ID Number Staff/1213/003
(Project: “The Hearing Body”).

For the following please circle “Yes” or “No” and initial each point.

_____ I agree for the videotape and photographs to be used by the researchers in further research studies YES / NO

_____ I agree for the videotape and photographs to be used by the researchers to demonstrate the technology YES / NO

_____ I agree for the videotape and photographs to be used by the researchers for teaching, conferences, presentations,
publications and/or thesis work YES / NO

_____ I agree to be contacted in the future by UCL researchers who would like to invite me to participate in follow-up
studies YES / NO

_____ I understand that my participation will be taped/video recorded/photographed and I am aware of and consent to the
use of the recordings/photographs YES / NO

Interviewer signature: ____________________________________

Interviewee signature: ____________________________________ Consent date: ___/____/___

Address _______________________________________________________________________

Phone number: _______________

116
APPENDIX E – MATLAB SCRIPTS FOR THE EXTRACTION OF
THE PRESSURE SENSOR AND ACCELEROMETER DATA

sensor1=data(:,1);
sensor2=data(:,2);
sensor3=data(:,3);
sensor4=data(:,4);
acc1=data(:,5);
acc2=data(:,6);
acc3=data(:,7);

[ RESULT ] = read_sensor( sensor1, 300, 20);

[ RESULT ] = cat(1,[RESULT],read_sensor( sensor2, 300, 20));
[ RESULT ] = cat(1,[RESULT],read_sensor( sensor3, 300, 20));
[ RESULT ] = cat(1,[RESULT],read_sensor( sensor4, 300, 20));

[ RESULT_ACC ] = read_acc(acc1,-130,20);
[ RESULT_ACC ] = read_acc(acc2,760,20);
[ RESULT_ACC ] = read_acc(acc3,1030,20);

break
% % %check graph
start1=1303;
last1=1822;
start2=1330;
start3=1280;
last3=1775;
start_acc=1353;
last_acc=1822;

dt12= start2-start1;
dt23=start2-start3;

New_result = read_sensor( sensor1(start1:last1), 300, 20);

final_sensor1= [];
for i=1 : length(New_result)
v=New_result(i,2);
t=New_result(i,1);

[t0, t1, avg]=feature_extraction(t,sensor1);

final_sensor1=[final_sensor1;[t0+(start2-start1) t1+(start2-start1) v t+(start2-start1) avg]];
end

New_result = read_sensor( sensor2(start2:last1), 300, 20);

%[New_result] = [New_result;[read_sensor( sensor2(start1:last1), 300, 20)]];

final_sensor2= [];

for i=1 : length(New_result)

v=New_result(i,2);
t=New_result(i,1);
[t0, t1, avg]=feature_extraction(t,sensor2);
final_sensor2=[final_sensor2;[t0 t1 v t avg]];
end

117
New_result = read_sensor( sensor3(start3:last3), 300, 20);

final_sensor3= [];
for i=1 : length(New_result)
v=New_result(i,2);
t=New_result(i,1);

[t0, t1, avg]=feature_extraction(t,sensor3);

final_sensor3=[final_sensor3;[t0+(start2-start3) t1+(start2-start3) v t+(start2-start3) avg]];
end

New_result = read_sensor( sensor4(start3:last3), 630, 20);

final_sensor4= [];
for i=1 : length(New_result)
v=New_result(i,2);
t=New_result(i,1);

[t0, t1, avg]=feature_extraction(t,sensor4);

final_sensor4=[final_sensor4;[t0+(start2-start3) t1+(start2-start3) v t+(start2-start3) avg]];
end

New_result_acc1 = read_acc( acc1(start_acc:last_acc), -130, 20);

final_acc1=[];
for i=1:length(New_result_acc1)
t=New_result_acc1(i,1);
v=New_result_acc1(i,2);

[t_min, v_min]=min_acc_extraction(t,v, acc1);

final_acc1=[final_acc1;[t_min, v_min]];
end

New_result_acc2 = read_acc( acc2(start_acc:last_acc), 760, 20);

final_acc2=[];
for i=1:length(New_result_acc)
t=New_result_acc(i,1);
v=New_result_acc(i,2);
[t_min, v_min]=min_acc_extraction(t,v, acc2);
final_acc2=[final_acc2;[t_min, v_min]];
end

New_result_acc3 = read_acc( acc3(start_acc:last_acc), 1030, 20);

final_acc3=[];

for i=1:length(New_result_acc)
t=New_result_acc(i,1);
v=New_result_acc(i,2);
[t_min, v_min]=min_acc_extraction(t,v,acc3);
final_acc3=[final_acc3;[t_min, v_min]];
end
min_acc_all_axis_array= [final_acc1 final_acc2 final_acc3];
max_acc_all_axis_array= [New_result_acc1 New_result_acc2 New_result_acc3];
all_sensor_array= [final_sensor1 final_sensor2 final_sensor3 final_sensor4];

118
APPENDIX F – MATLAB SCRIPTS FOR THE EXTRACTION OF
THE GSR DATA
close all
markers=P01_markers_time;
GSR=data(:,7);
% data=P01timestamp;
final_values=[];
N = length(GSR);
Fs=32;
t = linspace(0, N/Fs, N);
t=t';
for i=1:3:length(markers)-2

march_time=find(data(:,1)==markers(i,2))
walk_time=find(data(:,1)==markers(i+1,2))
end_time=find(data(:,1)==markers(i+2,2))

time_block=t(march_time:end_time)-t(march_time);

figure
block=[time_block GSR(march_time:end_time)];
plot(time_block,GSR(march_time:end_time),'c'), hold on
GSR_filter=sgolayfilt(GSR(march_time:end_time),1,13);
plot(time_block,GSR_filter, 'r');

baseline=sgolayfilt(GSR(march_time-Fs*1:march_time,:),1,13); %baseline
march=sgolayfilt(GSR(march_time:walk_time,:),1,13);
walk=sgolayfilt(GSR(walk_time:end_time),1,13);

avg_baseline=mean(baseline);
avg_march=mean(march);
avg_walk=mean(walk);
v1=avg_march-avg_baseline;
v2=avg_walk-avg_baseline;
v3=max(march)-min(march);
v4=max(walk)-min(walk);

final_values=[final_values;[v1 v2 v3 v4]];

end

119
APPENDIX G

Questionnaires
Perceived body dimensions/weight – 3D Body Visualizing Application

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

*
NF_R1_Body .099 22 .200 .964 22 .582
NF2_R1_Body .153 22 .194 .956 22 .408
*
LF_R1_Body .115 22 .200 .977 22 .868
*
LF2_R1_Body .115 22 .200 .976 22 .847
*
HF_R1_Body .116 22 .200 .980 22 .916
*
HF2_R1_Body .093 22 .200 .957 22 .423
Table 1: Test of normality for perceived body dimensions normalised values with the
Reciprocal Transformation (1/Χi).

Source df F Sig.

sound Sphericity Assumed 2 3.584 .037

Greenhouse-Geisser 1.614 3.584 .048

Huynh-Feldt 1.728 3.584 .044

Lower-bound 1.000 3.584 .072

Error(sound) Sphericity Assumed 42
Greenhouse-Geisser 33.893
Huynh-Feldt 36.296
Lower-bound 21.000
repetition Sphericity Assumed 1 .161 .692
Greenhouse-Geisser 1.000 .161 .692
Huynh-Feldt 1.000 .161 .692
Lower-bound 1.000 .161 .692
Error(repetition) Sphericity Assumed 21
Greenhouse-Geisser 21.000
Huynh-Feldt 21.000
Lower-bound 21.000
sound * repetition Sphericity Assumed 2 .232 .794
Greenhouse-Geisser 1.580 .232 .742
Huynh-Feldt 1.688 .232 .757
Lower-bound 1.000 .232 .635
Error(sound*repetition) Sphericity Assumed 42
Greenhouse-Geisser 33.190

120
 Huynh-Feldt 35.438

Lower-bound 21.000
Table 2: ANOVA for the detection of significant sound effects and interaction

M SE
NF_mean 57.5227 3.02523
LF_mena 56.7727 2.88351
HF_mean 54.6591 2.72005
Table 3: Means and Standard Errors for perceived body dimensions - non-transformed
values

T
df Sig. (2-tailed)

Pair 1 NF_R1_Body_mean -
-.547 21 .590
LF_R1_Body_mean
Pair 2 NF_R1_Body_mean -
-2.414 21 .025
HF_R1_Body_mean
Pair 3 LF_R1_Body_mean -
-1.775 21 .090
HF_R1_Body_mean
Table 4: t-test for the detection of sound effects between specific conditions

Task Experience Questionnaire

Shapiro-Wilk

Statistic df Sig.

NF_slow_quick .814 22 .001

NF_light_heavy .322 22 .000
NF_weak_strong .882 22 .013
NF_stoopedhatched_straight .911 22 .050
NF_soundswereproducedbymy
.751 22 .000
ownfootsteps
NF_thefeelingofmybodywasles
.920 22 .075
svividthannormal

NF_thefeelingsaboutmybodyw
.923 22 .090
eresurprisingandunexpected
NF_Icouldreallytellwheremyfe
.907 22 .041
etwere

121
NF2_slow_quick .882 22 .013
NF2_light_heavy .913 22 .053
NF2_weak_strong .904 22 .036
NF2_stoopedhatched_straight .892 22 .021
NF2_soundswereproducedbym
.745 22 .000
yownfootsteps
NF2_thefeelingofmybodywasle
.895 22 .024
ssvividthannormal
NF2_thefeelingsaboutmybody
.919 22 .072
weresurprisingandunexpected
NF2_Icouldreallytellwheremyf
.917 22 .065
eetwere
LF_slow_quick .884 22 .015
LF_light_heavy .935 22 .156
LF_weak_strong .859 22 .005
LF_stoopedhatched_straight .887 22 .017
LF_soundswereproducedbymy
.714 22 .000
ownfootsteps
LF_thefeelingofmybodywasles
.850 22 .003
svividthannormal
LF_thefeelingsaboutmybodywe
.937 22 .168
resurprisingandunexpected
LF_Icouldreallytellwheremyfee
.882 22 .013
twere
LF2_slow_quick .923 22 .088
LF2_light_heavy .946 22 .264
LF2_weak_strong .876 22 .010
LF2_stoopedhatched_straight .934 22 .149
LF2_soundswereproducedbym
.696 22 .000
yownfootsteps
LF2_thefeelingofmybodywasle
.899 22 .028
ssvividthannormal
LF2_thefeelingsaboutmybodyw
.929 22 .115
eresurprisingandunexpected
LF2_Icouldreallytellwheremyfe
.872 22 .009
etwere
HF_slow_quick .898 22 .027
HF_light_heavy .938 22 .178
HF_weak_strong .807 22 .001
HF_stoopedhatched_straight .932 22 .132
HF_soundswereproducedbymy
.646 22 .000
ownfootsteps

122
HF_thefeelingofmybodywasles
.899 22 .029
svividthannormal
HF_thefeelingsaboutmybodyw
.932 22 .135
eresurprisingandunexpected
HF_Icouldreallytellwheremyfe
.848 22 .003
etwere
HF2_slow_quick .890 22 .018
HF2_light_heavy .937 22 .175
HF2_weak_strong .870 22 .008
HF2_stoopedhatched_straight .843 22 .003
HF2_soundswereproducedbym
.592 22 .000
yownfootsteps
HF2_thefeelingofmybodywasle
.853 22 .004
ssvividthannormal
HF2_thefeelingsaboutmybody
.927 22 .106
weresurprisingandunexpected
HF2_Icouldreallytellwheremyf
.849 22 .003
eetwere
Table 5: Shapiro-Wilk test of normality for all the task experience questions

Perceived speed - Slow-Quick scale

NF2_slow_quick LF2_slow_quick - HF2_slow_quick

- NF_slow_quick LF_slow_quick - HF_slow_quick

Z -1.214b -1.518b -1.642b

Asymp. Sig. (2-tailed) .225 .129 .101

Table 6: Wilcoxon test within each condition and its repetition

N 22
Chi-Square 8.361
df 2
Asymp. Sig. .015
Table 7: Friedman’s ANOVA for the detection of significant sound effects

LF_mean_slow_q HF_mean_slow_q HF_mean_slow_q

uick - uick - uick -
NF_mean_slow_q NF_mean_slow_q LF_mean_slow_q
uick uick uick
b c
Z -.747 -2.035 -2.543c
Asymp. Sig. (2-tailed) .455 .042 .011
Table 8: Wilcoxon test within the three different sound conditions

123
Perceived weight – Light-Heavy scale

NF2_light_heavy LF2_light_heavy HF2_light_heavy

- NF_light_heavy - LF_light_heavy - HF_light_heavy
b b
Z -1.645 -1.603 -.765b
Asymp. Sig. (2-tailed) .100 .109 .444
Table 9: Wilcoxon test within each condition and its repetition

N 22
Chi-Square 4.200
df 2
Asymp. Sig. .122
Table 10: Friedman’s ANOVA

LF_mean_light_h HF_mean_light_h HF_mean_light_h

eavy - eavy - eavy -
NF_mean_light_h NF_mean_light_h LF_mean_light_h
eavy eavy eavy

Z -.631b -1.712c -2.020c

Asymp. Sig. (2-tailed) .528 .087 .043
Table 11: Wilcoxon test within the three different sound conditions

Perceived strength – Weak-Strong scale

NF2_weak_stron HF2_weak_stron
g- LF2_weak_strong g-
NF_weak_strong - LF_weak_strong HF_weak_strong
b c
Z -.492 -.731 -1.934b
Asymp. Sig. (2-tailed) .623 .465 .053
Table 12: Wilcoxon test within each condition and its repetition

N 22
Chi-Square 3.757
df 2
Asymp. Sig. .153
Table 13: Friedman’s ANOVA

LF_mean_weak_s HF_mean_weak_ HF_mean_weak_

trong - strong - strong -
NF_mean_weak_ NF_mean_weak_ LF_mean_weak_s
strong strong trong
b c
Z -.439 -1.559 -1.805c

124
Asymp. Sig. (2-tailed) .660 .119 .071
Table 14: Wilcoxon test within the three different sound conditions

Perceived posture - Stooped/Hatched – Straight scale

NF2_stoopedhatc LF2_stoopedhatc HF2_stoopedhatc

hed_straight - hed_straight - hed_straight -
NF_stoopedhatch LF_stoopedhatche HF_stoopedhatch
ed_straight d_straight ed_straight

Z -.322b -1.344c -2.003c

Asymp. Sig. (2-tailed) .747 .179 .045
Table 15: Wilcoxon test within each condition and its repetition

N 22
Chi-Square 3.836
df 2
Asymp. Sig. .147
Table 16: Friedman’s ANOVA - repetition 2

LF2_stoopedhatc HF2_stoopedhatc HF2_stoopedhatc

hed_straight - hed_straight - hed_straight -
NF2_stoopedhatc NF2_stoopedhatc LF2_stoopedhatc
hed_straight hed_straight hed_straight
b b
Z -.914 -2.094 -1.532b
Asymp. Sig. (2-tailed) .361 .036 .125
Table 17: Wilcoxon test within the three different sound conditions - repetition 2

Identification of foot location

NF2_identificatio LF2_identificatio HF2_identificatio

n_of_foot_locatio n_of_foot_locatio n_of_foot_locatio
n n n
NF_identification LF_identification HF_identification
_of_foot_location _of_foot_location _of_foot_location
b
Z -.072 -.527b -.664b
Asymp. Sig. (2-tailed) .943 .598 .507
Table 18: Wilcoxon for exploration of sound repetition effect

N 22
Chi-Square 8.517
df 2
Asymp. Sig. .014
Table 19: Friedman’s ANOVA

125
 LF_identification HF_identification HF_identification
_of_foot_location _of_foot_location _of_foot_location
_mean _mean _mean
NF_identification NF_identification LF_identification
_of_foot_location _of_foot_location _of_foot_location
_mean _mean _mean
b b
Z -.095 -2.293 -2.063b
Asymp. Sig. (2-tailed) .924 .022 .039
Table 20: Wilcoxon test within the three different sound conditions

Participants’ ability to identify their own feet as the footsteps sound source

NF2_sounds_sour LF2_sounds_sour HF2_sounds_sour

ce ce - ce
NF_sounds_sourc LF_sounds_sourc HF_sounds_sourc
e e e

Z -.137b -.577b -.302c

Asymp. Sig. (2-tailed) .891 .564 .763
Table 21: Wilcoxon between each condition and its repetition

N 22
Chi-Square .241
df 2
Asymp. Sig. .886
Table 22: Friedman’s ANOVA

Experience of surprising and unexpected feelings of body

NF2_surprising.u LF2_surprising.u HF2_surprising.u
nexpected_feeling nexpected_feeling nexpected_feeling
s s s
NF_surprising.un LF_surprising.une HF_surprising.un
expected_feelings xpected_feelings expected_feelings
b c
Z -.218 -.835 -.966c
Asymp. Sig. (2-tailed) .827 .404 .334
Table 23: Wilcoxon between the two repetitions of each condition

Source df F Sig.
sound Sphericity Assumed 2 .284 .754

126
 Greenhouse-Geisser 1.645 .284 .712

Huynh-Feldt 1.767 .284 .728

Lower-bound 1.000 .284 .600

Error(sound) Sphericity Assumed 42
Greenhouse-Geisser 34.551
Huynh-Feldt 37.104
Lower-bound 21.000
repetition Sphericity Assumed 1 .468 .502
Greenhouse-Geisser 1.000 .468 .502
Huynh-Feldt 1.000 .468 .502
Lower-bound 1.000 .468 .502
Error(repetition) Sphericity Assumed 21
Greenhouse-Geisser 21.000
Huynh-Feldt 21.000
Lower-bound 21.000
sound * repetition Sphericity Assumed 2 .571 .569
Greenhouse-Geisser 1.931 .571 .564
Huynh-Feldt 2.000 .571 .569
Lower-bound 1.000 .571 .458
Error(sound*repetition) Sphericity Assumed 42

Greenhouse-Geisser 40.554

Huynh-Feldt 42.000

Lower-bound 21.000
Table 24: ANOVA

Vividness of perceived body feelings.

NF2__vividness_
of_body_feelings LF2_vividness_of HF2_vividness_o
- _body_feelings - f_body_feelings -
NF_vividness_of LF_vividness_of_ HF_vividness_of
_body_feelings body_feelings _body_feelings

Z -1.509b -1.530c -1.275b

Asymp. Sig. (2-tailed) .131 .126 .202
Table 25: Wilcoxon between each condition and its repetition

N 22
Chi-Square .200
df 2
Asymp. Sig. .905
Table 26: Friedman’s ANOVA

Self-Assessment Manikin (SAM)

127
 Shapiro-Wilk

Statistic df Sig.

NF_SAM_Unhappy_Happy .925 22 .096

NF_SAM_Unaroused_aroused .936 22 .167
NF_SAM_Submissive_domina
.927 22 .104
nt
NF2_SAM_Unhappy_Happy .913 22 .055
NF2_SAM_Unaroused_arouse
.930 22 .125
d
NF2_SAM_Submissive_domin
.935 22 .154
ant
LF_SAM_Unhappy_Happy .939 22 .185
LF_SAM_Unaroused_aroused .946 22 .265
LF_SAM_Submissive_domina
.931 22 .131
nt
LF2_SAM_Unhappy_Happy .923 22 .088
LF2_SAM_Unaroused_aroused .954 22 .377
LF2_SAM_Submissive_domin
.963 22 .557
ant
HF_SAM_Unhappy_Happy .883 22 .014
HF_SAM_Unaroused_aroused .939 22 .190
HF_SAM_Submissive_domina
.874 22 .009
nt
HF2_SAM_Unhappy_Happy .905 22 .037
HF2_SAM_Unaroused_arouse
.907 22 .042
d
Table 27: Shapiro-Wilk test of normality for all SAM variables

SAM – Dominance

NF2_SAM_Domi LF2_SAM_Domi HF2_SAM_Domi

nance nance nance
NF_SAM_Domin LF_SAM_Domin HF_SAM_Domin
ance ance ance
b c
Z -.515 -.656 -1.442c
Asymp. Sig. (2-tailed) .607 .512 .149
Table 28: Wilcoxon for exploration of sound repetition effect

N 22
Chi-Square 2.471
df 2
Asymp. Sig. .291

128
Table 29: Friedman’s ANOVA

LF_mean_SAM_ HF_mean_SAM_ HF_mean_SAM_

Submissive_domi Submissive_domi Submissive_domi
nant - nant - nant -
NF_mean_SAM_ NF_mean_SAM_ LF_mean_SAM_
Submissive_domi Submissive_domi Submissive_domi
nant nant nant

Z -.088b -1.748c -2.006c

Asymp. Sig. (2-tailed) .930 .080 .045
Table 30: Wilcoxon test within the three different sound conditions

SAM – Valence

LF2_SAM_Valen HF2_SAM_Valen
ce- ce-
NF2_SAM_Valen LF_SAM_Valenc HF_SAM_Valenc
ce- NF_Valence e- e-
b b
Z -.695 -1.203 -1.567b
Asymp. Sig. (2-tailed) .487 .229 .117
Table 31: Wilcoxon for exploration of sound repetition effect

N 22
Chi-Square 6.222
df 2
Asymp. Sig. .045
Table 32: Friedman’s ANOVA

LF_ HF_ HF_

Valence_mean Valence_mean Valence_mean
NF_ NF_ LF_
Valence_mean Valence_mean Valence_mean
b b
Z -.063 -2.395 -1.819b
Asymp. Sig. (2-tailed) .950 .017 .069
Table 33: Wilcoxon test within the three different sound conditions

SAM – Arousal

NF2_SAM_Unar LF2_SAM_Unaro HF2_SAM_Unar

oused_aroused - used_aroused - oused_aroused -
NF_SAM_Unaro LF_SAM_Unarou HF_SAM_Unaro
used_aroused sed_aroused used_aroused

Z -.500b -.217c -1.778c

Asymp. Sig. (2-tailed) .617 .828 .075

129
Table 34: Wilcoxon for exploration of sound repetition effect

N 22
Chi-Square 2.324
df 2
Asymp. Sig. .313
Table 35: Friedman’s ANOVA

Spanner

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

NF_R1_Spanner_W_z .103 22 .200* .976 22 .836

NF_R1_Spanner_L_z .168 22 .108 .953 22 .369
NF2_R1_Spanner_W_z .188 22 .041 .929 22 .117
*
NF2_R1_Spanner_L_z .151 22 .200 .950 22 .312
*
LF_R1_Spanner_W_z .145 22 .200 .950 22 .319
*
LF_R1_Spanner_L_z .129 22 .200 .955 22 .402
*
LF2_R1_Spanner_W_z .118 22 .200 .949 22 .300
LF2_R1_Spanner_L_z .187 22 .044 .953 22 .361
*
HF_R1_Spanner_W_z .111 22 .200 .938 22 .179
HF_R1_Spanner_L_z .160 22 .147 .934 22 .146
*
HF2_R1_Spanner_W_z .094 22 .200 .978 22 .878
*
HF2_R1_Spanner_L_z .128 22 .200 .966 22 .619
Table 36: Shapiro-Wilk test of normality for both weight and length values in each
experimental trial

Within Subjects Effect Value F Hypothesis df Error df Sig.

sound Pillai's Trace .052 .564 4.000 84.000 .690

Wilks' Lambda .948 .557c 4.000 82.000 .694

Hotelling's Trace .055 .550 4.000 80.000 .699

Roy's Largest Root .054 1.128d 2.000 42.000 .333

c
repetition Pillai's Trace .034 .356 2.000 20.000 .705
c
Wilks' Lambda .966 .356 2.000 20.000 .705
Hotelling's Trace .036 .356c 2.000 20.000 .705
c
Roy's Largest Root .036 .356 2.000 20.000 .705
sound * repetition Pillai's Trace .106 1.175 4.000 84.000 .328

Wilks' Lambda .896 1.163c 4.000 82.000 .333

Hotelling's Trace .115 1.150 4.000 80.000 .339

Roy's Largest Root .098 2.056d 2.000 42.000 .141

130
Table 37: Multivariate test to detect sound effects and interaction

Source Measure df F Sig.

sound w Sphericity Assumed 2 1.089 .346

Greenhouse-Geisser 1.989 1.089 .345

Huynh-Feldt 2.000 1.089 .346

Lower-bound 1.000 1.089 .308

l Sphericity Assumed 2 .107 .899

Greenhouse-Geisser 1.621 .107 .859

Huynh-Feldt 1.737 .107 .873

Lower-bound 1.000 .107 .747

Error(sound) w Sphericity Assumed 42
Greenhouse-Geisser 41.772
Huynh-Feldt 42.000
Lower-bound 21.000
l Sphericity Assumed 42
Greenhouse-Geisser 34.034
Huynh-Feldt 36.469
Lower-bound 21.000
repetition w Sphericity Assumed 1 .264 .613
Greenhouse-Geisser 1.000 .264 .613
Huynh-Feldt 1.000 .264 .613
Lower-bound 1.000 .264 .613
l Sphericity Assumed 1 .602 .446
Greenhouse-Geisser 1.000 .602 .446
Huynh-Feldt 1.000 .602 .446
Lower-bound 1.000 .602 .446
Error(repetition) w Sphericity Assumed 21
Greenhouse-Geisser 21.000
Huynh-Feldt 21.000
Lower-bound 21.000
l Sphericity Assumed 21
Greenhouse-Geisser 21.000
Huynh-Feldt 21.000
Lower-bound 21.000
sound * repetition w Sphericity Assumed 2 .871 .426
Greenhouse-Geisser 1.932 .871 .423
Huynh-Feldt 2.000 .871 .426
Lower-bound 1.000 .871 .361
l Sphericity Assumed 2 1.223 .305

131
 Greenhouse-Geisser 1.829 1.223 .303
Huynh-Feldt 1.995 1.223 .305
Lower-bound 1.000 1.223 .281
Error(sound*repetition) w Sphericity Assumed 42

Greenhouse-Geisser 40.570

Huynh-Feldt 42.000

Lower-bound 21.000

l Sphericity Assumed 42

Greenhouse-Geisser 38.408

Huynh-Feldt 41.885

Lower-bound 21.000
Table 38: Univariate test to detect possible effect between the weight and length variables

Self-Efficacy

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

*
perc_str__NF_mean .152 22 .200 .964 22 .582
*
perc_str__NF2_mean .126 22 .200 .947 22 .274
perc_str__LF_mean .166 22 .120 .927 22 .105
perc_str__LF2_mean .162 22 .140 .957 22 .428
*
perc_str__HF_mean .107 22 .200 .965 22 .595
*
perc_str__HF2_mean .102 22 .200 .981 22 .925
Table 39: Shapiro-Wilk test of normality

Source df F Sig.

sound Sphericity Assumed 2 1.629 .208

Greenhouse-Geisser 1.852 1.629 .210

Huynh-Feldt 2.000 1.629 .208

Lower-bound 1.000 1.629 .216

Error(sound) Sphericity Assumed 42
Greenhouse-Geisser 38.900
Huynh-Feldt 42.000
Lower-bound 21.000
repetition Sphericity Assumed 1 .814 .377
Greenhouse-Geisser 1.000 .814 .377
Huynh-Feldt 1.000 .814 .377
Lower-bound 1.000 .814 .377
Error(repetition) Sphericity Assumed 21
Greenhouse-Geisser 21.000
Huynh-Feldt 21.000

132
 Lower-bound 21.000
sound * repetition Sphericity Assumed 2 3.698 .033
Greenhouse-Geisser 1.974 3.698 .034
Huynh-Feldt 2.000 3.698 .033
Lower-bound 1.000 3.698 .068
Error(sound*repetition) Sphericity Assumed 42

Greenhouse-Geisser 41.457

Huynh-Feldt 42.000

Lower-bound 21.000
Table 40: sound x repetition ANOVA

Behavioural measures
Time of heel contact with the ground

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

HF1_Right_heel_t1t0diff .122 16 .200* .917 16 .152

*
HF2_Right_heel_t1t0diff .124 16 .200 .958 16 .632
*
LF1_Right_heel_t1t0diff .135 16 .200 .947 16 .446
*
LF2_Right_heel_t1t0diff .147 16 .200 .930 16 .247
*
NF1_Right_heel_t1t0diff .167 16 .200 .970 16 .835
*
NF2_Right_heel_t1t0diff .098 16 .200 .979 16 .957
*
HF1_Left_heel_t1t0diff .135 16 .200 .963 16 .717
*
HF2_Left_heel_t1t0diff .150 16 .200 .950 16 .490
*
LF1_Left_heel_t1t0diff .118 16 .200 .956 16 .594
LF2_Left_heel_t1t0diff .193 16 .115 .953 16 .542
*
NF1_Left_heel_t1t0diff .153 16 .200 .955 16 .572
*
NF2_Left_heel_t1t0diff .136 16 .200 .958 16 .618
Table 41: Shapiro-Wilk test of normality

Source df F Sig.

foot Sphericity Assumed 1 2.085 .169

Greenhouse-Geisser 1.000 2.085 .169

Huynh-Feldt 1.000 2.085 .169

Lower-bound 1.000 2.085 .169

Error(foot) Sphericity Assumed 15
Greenhouse-Geisser 15.000
Huynh-Feldt 15.000
Lower-bound 15.000
sound Sphericity Assumed 2 2.534 .096
Greenhouse-Geisser 1.692 2.534 .107

133
 Huynh-Feldt 1.884 2.534 .100
Lower-bound 1.000 2.534 .132
Error(sound) Sphericity Assumed 30
Greenhouse-Geisser 25.381
Huynh-Feldt 28.261
Lower-bound 15.000
repetition Sphericity Assumed 1 .002 .965
Greenhouse-Geisser 1.000 .002 .965
Huynh-Feldt 1.000 .002 .965
Lower-bound 1.000 .002 .965
Error(repetition) Sphericity Assumed 15
Greenhouse-Geisser 15.000
Huynh-Feldt 15.000
Lower-bound 15.000
foot * sound Sphericity Assumed 2 .766 .474
Greenhouse-Geisser 1.758 .766 .459
Huynh-Feldt 1.973 .766 .472
Lower-bound 1.000 .766 .395
Error(foot*sound) Sphericity Assumed 30
Greenhouse-Geisser 26.368
Huynh-Feldt 29.594
Lower-bound 15.000
foot * repetition Sphericity Assumed 1 .163 .692
Greenhouse-Geisser 1.000 .163 .692
Huynh-Feldt 1.000 .163 .692
Lower-bound 1.000 .163 .692
Error(foot*repetition) Sphericity Assumed 15
Greenhouse-Geisser 15.000
Huynh-Feldt 15.000
Lower-bound 15.000
sound * repetition Sphericity Assumed 2 .519 .601
Greenhouse-Geisser 1.639 .519 .566
Huynh-Feldt 1.812 .519 .584
Lower-bound 1.000 .519 .483
Error(sound*repetition) Sphericity Assumed 30
Greenhouse-Geisser 24.578
Huynh-Feldt 27.187
Lower-bound 15.000
foot * sound * repetition Sphericity Assumed 2 3.123 .059
Greenhouse-Geisser 1.570 3.123 .073
Huynh-Feldt 1.722 3.123 .068

134
 Lower-bound 1.000 3.123 .098
Error(foot*sound*repetition) Sphericity Assumed 30

Greenhouse-Geisser 23.557

Huynh-Feldt 25.832

Lower-bound 15.000
Table 42: foot x sound x repetition ANOVA

Source df F Sig.

sound Sphericity Assumed 2 3.956 .029

Greenhouse-Geisser 1.376 3.956 .047

Huynh-Feldt 1.457 3.956 .044

Lower-bound 1.000 3.956 .063

Error(sound) Sphericity Assumed 34
Greenhouse-Geisser 23.386
Huynh-Feldt 24.765
Lower-bound 17.000
repetition Sphericity Assumed 1 .002 .962
Greenhouse-Geisser 1.000 .002 .962
Huynh-Feldt 1.000 .002 .962
Lower-bound 1.000 .002 .962
Error(repetition) Sphericity Assumed 17
Greenhouse-Geisser 17.000
Huynh-Feldt 17.000
Lower-bound 17.000
sound * repetition Sphericity Assumed 2 .275 .761
Greenhouse-Geisser 1.770 .275 .735
Huynh-Feldt 1.960 .275 .757
Lower-bound 1.000 .275 .607
Error(sound*repetition) Sphericity Assumed 34

Greenhouse-Geisser 30.082

Huynh-Feldt 33.319

Lower-bound 17.000
Table 43: Left foot sensor ANOVA

t df Sig. (2-tailed)

Pair 1 HF_Left_heel_t1t0diff_mean -
-2.070 21 .051
LF_Left_heel_t1t0diff_mean

135
Pair 2 HF_Left_heel_t1t0diff_mean -
-.686 21 .500
NF_Left_heel_t1t0diff_mean
Pair 3 LF_Left_heel_t1t0diff_mean -
2.033 21 .055
NF_Left_heel_t1t0diff_mean
Table 44: Paired sample t-test

Average of toe values

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

*
HF1_R_toe_avg_z .130 16 .200 .952 16 .518
*
HF2_R_toe_avg_z .165 16 .200 .906 16 .102
*
LF1_R_toe_avg_z .151 16 .200 .909 16 .113
*
LF2_R_toe_avg_z .124 16 .200 .935 16 .296
*
NF1_R_toe_avg_z .131 16 .200 .945 16 .415
*
NF2_R_toe_avg_z .102 16 .200 .950 16 .485
*
HF1_L_toe_avg_z .102 16 .200 .968 16 .803
*
HF2_L_toe_avg_z .126 16 .200 .951 16 .507
*
LF1_L_toe_avg_z .098 16 .200 .952 16 .529
*
LF2_L_toe_avg_z .147 16 .200 .928 16 .230
NF1_L_toe_avg_z .190 16 .126 .953 16 .534
*
NF2_L_toe_avg_z .147 16 .200 .964 16 .735
Table 45: Shapiro-Wilk test for both left and right feet sensors

Source df F Sig.

sound Sphericity Assumed 2 4.518 .018

Greenhouse-Geisser 1.790 4.518 .022

Huynh-Feldt 1.987 4.518 .018

Lower-bound 1.000 4.518 .048

Error(sound) Sphericity Assumed 34
Greenhouse-Geisser 30.430
Huynh-Feldt 33.776
Lower-bound 17.000
repetition Sphericity Assumed 1 .327 .575
Greenhouse-Geisser 1.000 .327 .575
Huynh-Feldt 1.000 .327 .575
Lower-bound 1.000 .327 .575
Error(repetition) Sphericity Assumed 17
Greenhouse-Geisser 17.000
Huynh-Feldt 17.000
Lower-bound 17.000
sound * repetition Sphericity Assumed 2 2.573 .091

136
 Greenhouse-Geisser 1.970 2.573 .092
Huynh-Feldt 2.000 2.573 .091
Lower-bound 1.000 2.573 .127
Error(sound*repetition) Sphericity Assumed 34

Greenhouse-Geisser 33.487

Huynh-Feldt 34.000

Lower-bound 17.000
Table 46: ANOVA – sound by repetition

t df Sig. (2-tailed)

Pair 1 HF_L_toe_avg_z_mean -
-2.217 21 .038
LF_L_toe_avg_z_mean
Pair 2 HF_L_toe_avg_z_mean -
-.274 21 .787
NF_L_toe_avg_z_mean
Pair 3 LF_L_toe_avg_z_mean -
1.613 21 .122
NF_L_toe_avg_z_mean
Table 47: Paired t-test for left foot sensors

Average step time – time interval between heel’s initial strike and toe’s last contact with
the ground

Shapiro-Wilk

Statistic df Sig.

HF1_Right_heel_to_toe .599 16 .000

HF2_Right_heel_to_toe .927 16 .216
LF1_Right_heel_to_toe .571 16 .000
LF2_Right_heel_to_toe .954 16 .552
NF1_Right_heel_to_toe .633 16 .000
NF2_Right_heel_to_toe .935 16 .293
HF1_Left_heel_to_toe .961 16 .689
HF2_Left_heel_to_toe .957 16 .603
LF1_Left_heel_to_toe .971 16 .853
LF2_Left_heel_to_toe .947 16 .451
NF1_Left_heel_to_toe .954 16 .548
NF2_Left_heel_to_toe .978 16 .942
Table 48: Shapiro-Wilk test for both feet – normal distribution for the left foot sensor data

Source df F Sig.

sound Sphericity Assumed 2 .795 .460

Greenhouse-Geisser 1.791 .795 .448

Huynh-Feldt 1.988 .795 .459

137
 Lower-bound 1.000 .795 .385
Error(sound) Sphericity Assumed 34
Greenhouse-Geisser 30.440
Huynh-Feldt 33.790
Lower-bound 17.000
repetition Sphericity Assumed 1 .036 .852
Greenhouse-Geisser 1.000 .036 .852
Huynh-Feldt 1.000 .036 .852
Lower-bound 1.000 .036 .852
Error(repetition) Sphericity Assumed 17
Greenhouse-Geisser 17.000
Huynh-Feldt 17.000
Lower-bound 17.000
sound * repetition Sphericity Assumed 2 1.370 .268
Greenhouse-Geisser 1.519 1.370 .267
Huynh-Feldt 1.637 1.370 .267
Lower-bound 1.000 1.370 .258
Error(sound*repetition) Sphericity Assumed 34

Greenhouse-Geisser 25.829

Huynh-Feldt 27.836

Lower-bound 17.000
Table 49: ANOVA – sound x repetition – no significant sound effects or interaction

Average of the maximum heel pressure

Initial Shapiro-Wilk test showed non-normal distribution. Normalization was achieved by

converting the original values into z-scores. Therefore, z-scores were used for subsequent
analysis.

Shapiro-Wilk

Statistic df Sig.

HF1_Right_heel_peak_Z .962 16 .694

HF2_Right_heel_peak_Z .893 16 .061
LF1_Right_heel_peak_Z .947 16 .451
LF2_Right_heel_peak_Z .975 16 .911
NF1_Right_heel_peak_Z .927 16 .218
NF2_Right_heel_peak_Z .976 16 .920
HF1_Left_heel_peak_Z .939 16 .331
HF2_Left_heel_peak_Z .933 16 .276
LF1_Left_heel_peak_Z .951 16 .510
LF2_Left_heel_peak_Z .933 16 .268
NF1_Left_heel_peak_Z .974 16 .892

138
NF2_Left_heel_peak_Z .950 16 .487
Table 50: Shapiro wilk - z-scores

Source df F Sig.

sound Sphericity Assumed 2 .095 .910

Greenhouse-Geisser 1.675 .095 .878

Huynh-Feldt 1.837 .095 .895

Lower-bound 1.000 .095 .762

Error(sound) Sphericity Assumed 34
Greenhouse-Geisser 28.480
Huynh-Feldt 31.233
Lower-bound 17.000
repetition Sphericity Assumed 1 1.204 .288
Greenhouse-Geisser 1.000 1.204 .288
Huynh-Feldt 1.000 1.204 .288
Lower-bound 1.000 1.204 .288
Error(repetition) Sphericity Assumed 17
Greenhouse-Geisser 17.000
Huynh-Feldt 17.000
Lower-bound 17.000
sound * repetition Sphericity Assumed 2 .575 .568
Greenhouse-Geisser 1.701 .575 .542
Huynh-Feldt 1.871 .575 .557
Lower-bound 1.000 .575 .459
Error(sound*repetition) Sphericity Assumed 34

Greenhouse-Geisser 28.924

Huynh-Feldt 31.809

Lower-bound 17.000
Table 51: ANOVA showed no significant sound effects or interaction

Average of heel pressure applied on the ground

Tests of Normality
Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

HF1_Right_heel_average .149 16 .200* .897 16 .072

*
HF2_Right_heel_average .144 16 .200 .896 16 .069
LF1_Right_heel_average .224 16 .031 .867 16 .024
LF2_Right_heel_average .228 16 .026 .818 16 .005

139
NF1_Right_heel_average .147 16 .200* .914 16 .135
NF2_Right_heel_average .209 16 .061 .880 16 .038
*
HF1_Left_heel_average .138 16 .200 .954 16 .550
*
HF2_Left_heel_average .106 16 .200 .954 16 .548
*
LF1_Left_heel_average .113 16 .200 .984 16 .987
*
LF2_Left_heel_average .107 16 .200 .952 16 .528
NF1_Left_heel_average .196 16 .100 .922 16 .180
*
NF2_Left_heel_average .103 16 .200 .973 16 .890
Table 52: Shapiro-Wilk test for both left and right foot sensors – normal distribution in left
foot sensors

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

HF1_Right_heel_average .149 16 .200* .897 16 .072

*
HF2_Right_heel_average .144 16 .200 .896 16 .069
LF1_Right_heel_average .224 16 .031 .867 16 .024
LF2_Right_heel_average .228 16 .026 .818 16 .005
*
NF1_Right_heel_average .147 16 .200 .914 16 .135
NF2_Right_heel_average .209 16 .061 .880 16 .038
*
HF1_Left_heel_average .138 16 .200 .954 16 .550
*
HF2_Left_heel_average .106 16 .200 .954 16 .548
*
LF1_Left_heel_average .113 16 .200 .984 16 .987
*
LF2_Left_heel_average .107 16 .200 .952 16 .528
NF1_Left_heel_average .196 16 .100 .922 16 .180
*
NF2_Left_heel_average .103 16 .200 .973 16 .890
Table 53: ANOVA showed no significant effects or interaction

Average time of toe contact with the ground

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

HF1_Right_toe_t1t0diff .155 16 .200* .934 16 .284

*
HF2_Right_toe_t1t0diff .156 16 .200 .923 16 .191
*
LF1_Right_toe_t1t0diff .161 16 .200 .939 16 .338
*
LF2_Right_toe_t1t0diff .139 16 .200 .928 16 .227
*
NF1_Right_toe_t1t0diff .136 16 .200 .936 16 .301
*
NF2_Right_toe_t1t0diff .153 16 .200 .939 16 .339
*
HF1_Left_toe_t1t0diff .119 16 .200 .944 16 .396
*
HF2_Left_toe_t1t0diff .138 16 .200 .960 16 .671
LF1_Left_toe_t1t0diff .285 16 .001 .902 16 .086
*
LF2_Left_toe_t1t0diff .164 16 .200 .911 16 .123
*
NF1_Left_toe_t1t0diff .133 16 .200 .968 16 .805
*
NF2_Left_toe_t1t0diff .082 16 .200 .986 16 .993
Table 54: Shapiro-Wilk test for both left and right foot sensors – normal data distribution

140
Source df F Sig.

sound Sphericity Assumed 2 .277 .760

Greenhouse-Geisser 1.983 .277 .758

Huynh-Feldt 2.000 .277 .760

Lower-bound 1.000 .277 .605

Error(sound) Sphericity Assumed 34
Greenhouse-Geisser 33.714
Huynh-Feldt 34.000
Lower-bound 17.000
repetition Sphericity Assumed 1 .023 .882
Greenhouse-Geisser 1.000 .023 .882
Huynh-Feldt 1.000 .023 .882
Lower-bound 1.000 .023 .882
Error(repetition) Sphericity Assumed 17
Greenhouse-Geisser 17.000
Huynh-Feldt 17.000
Lower-bound 17.000
sound * repetition Sphericity Assumed 2 .169 .845
Greenhouse-Geisser 1.474 .169 .779
Huynh-Feldt 1.580 .169 .795
Lower-bound 1.000 .169 .686
Error(sound*repetition) Sphericity Assumed 34

Greenhouse-Geisser 25.053

Huynh-Feldt 26.855

Lower-bound 17.000
Table 55: ANOVA showed no significant effects or interaction

Average of maximum toe pressure

Initial Shapiro-Wilk test showed non-normal distribution. Normalization was achieved by

converting the original values into z-scores. Therefore, z-scores were used for subsequent
analysis.

Kolmogorov-Smirnova Shapiro-Wilk

Statistic df Sig. Statistic df Sig.

HF1_Right_toe_peak_z .138 16 .200* .972 16 .871

*
HF2_Right_toe_peak_z .124 16 .200 .948 16 .460
*
LF1_Right_toe_peak_z .154 16 .200 .939 16 .342
LF2_Right_toe_peak_z .181 16 .167 .898 16 .074
NF1_Right_toe_peak_z .199 16 .090 .887 16 .050
*
NF2_Right_toe_peak_z .140 16 .200 .942 16 .368
*
HF1_Left_toe_peak_z .153 16 .200 .941 16 .367

141
HF2_Left_toe_peak_z .139 16 .200* .950 16 .498
LF1_Left_toe_peak_z .192 16 .118 .959 16 .637
*
LF2_Left_toe_peak_z .134 16 .200 .962 16 .699
*
NF1_Left_toe_peak_z .111 16 .200 .955 16 .580
*
NF2_Left_toe_peak_z .129 16 .200 .944 16 .407
Table 56: Shapiro-Wilk - z-scores - test for both left and right foot sensors – normal
distribution in left foot sensors

Source df F Sig.

sound Sphericity Assumed 2 .007 .993

Greenhouse-Geisser 1.514 .007 .979

Huynh-Feldt 1.631 .007 .984

Lower-bound 1.000 .007 .933

Error(sound) Sphericity Assumed 34
Greenhouse-Geisser 25.738
Huynh-Feldt 27.721
Lower-bound 17.000
repetition Sphericity Assumed 1 .059 .811
Greenhouse-Geisser 1.000 .059 .811
Huynh-Feldt 1.000 .059 .811
Lower-bound 1.000 .059 .811
Error(repetition) Sphericity Assumed 17
Greenhouse-Geisser 17.000
Huynh-Feldt 17.000
Lower-bound 17.000
sound * repetition Sphericity Assumed 2 .103 .902
Greenhouse-Geisser 1.977 .103 .900
Huynh-Feldt 2.000 .103 .902
Lower-bound 1.000 .103 .752
Error(sound*repetition) Sphericity Assumed 34

Greenhouse-Geisser 33.613

Huynh-Feldt 34.000

Lower-bound 17.000

Table 57 15: ANOVA showed no significant effects or interaction

ACCELEROMETER

Kolmogorov-Smirnova Shapiro-Wilk

142
 Statistic df Sig. Statistic df Sig.

HF1_dec_lifting .128 19 .200* .951 19 .403

*
HF1_acc_down .093 19 .200 .957 19 .508
*
HF1_acc_lifting .097 19 .200 .954 19 .468
*
HF2_dec_lifting .121 19 .200 .952 19 .431
HF2_acc_down .165 19 .187 .923 19 .128
*
HF2_acc_lifting .134 19 .200 .965 19 .671
*
LF1_dec_lifting .148 19 .200 .928 19 .160
*
LF1_acc_down .141 19 .200 .961 19 .595
*
LF1_acc_lifting .132 19 .200 .971 19 .800
LF2_dec_lifting .164 19 .190 .925 19 .140
*
LF2_acc_down .107 19 .200 .986 19 .988
*
LF2_acc_lifting .103 19 .200 .959 19 .558
*
NF1_dec_lifting .131 19 .200 .943 19 .293
*
NF1_acc_down .101 19 .200 .969 19 .765
*
NF1_acc_lifting .090 19 .200 .971 19 .796
*
NF2_dec_lifting .109 19 .200 .975 19 .871
*
NF2_acc_down .146 19 .200 .962 19 .616
*
NF2_acc_lifting .090 19 .200 .965 19 .680
Table 58: Shapiro-Wilk test of normality for all accelerometer variables

Acceleration when lifting the foot to move forward

Source df F Sig.

sound Sphericity Assumed 2 1.941 .158

Greenhouse-Geisser 1.968 1.941 .159

Huynh-Feldt 2.000 1.941 .158

Lower-bound 1.000 1.941 .180

Error(sound) Sphericity Assumed 36
Greenhouse-Geisser 35.426
Huynh-Feldt 36.000
Lower-bound 18.000
repetition Sphericity Assumed 1 .651 .430
Greenhouse-Geisser 1.000 .651 .430
Huynh-Feldt 1.000 .651 .430
Lower-bound 1.000 .651 .430
Error(repetition) Sphericity Assumed 18
Greenhouse-Geisser 18.000
Huynh-Feldt 18.000
Lower-bound 18.000
sound * repetition Sphericity Assumed 2 .707 .500
Greenhouse-Geisser 1.889 .707 .493

143
 Huynh-Feldt 2.000 .707 .500
Lower-bound 1.000 .707 .412
Error(sound*repetition) Sphericity Assumed 36

Greenhouse-Geisser 33.995

Huynh-Feldt 36.000

Lower-bound 18.000
Table 59: ANOVA - sound x repetition

t df Sig.

Pair 1 HF_acc_lifting - LF_acc_lifting 2.301 21 .032

Pair 2 HF_acc_lifting - NF_acc_lifting 1.546 21 .137
Pair 3 LF_acc_lifting - NF_acc_lifting -.577 21 .570
Table 60: paired sample t-test

Acceleration when moving the foot downwards towards the floor.

Source df F Sig.

sound Sphericity Assumed 2 .045 .956

Greenhouse-Geisser 1.797 .045 .943

Huynh-Feldt 1.984 .045 .955

Lower-bound 1.000 .045 .833

Error(sound) Sphericity Assumed 36
Greenhouse-Geisser 32.344
Huynh-Feldt 35.706
Lower-bound 18.000
repetition Sphericity Assumed 1 3.854 .065
Greenhouse-Geisser 1.000 3.854 .065
Huynh-Feldt 1.000 3.854 .065
Lower-bound 1.000 3.854 .065
Error(repetition) Sphericity Assumed 18
Greenhouse-Geisser 18.000
Huynh-Feldt 18.000
Lower-bound 18.000
sound * repetition Sphericity Assumed 2 .810 .453
Greenhouse-Geisser 1.940 .810 .450
Huynh-Feldt 2.000 .810 .453
Lower-bound 1.000 .810 .380
Error(sound*repetition) Sphericity Assumed 36

Greenhouse-Geisser 34.916

Huynh-Feldt 36.000
Lower-bound 18.000

144
Table 61: ANOVA – sound x repetition – no significant sound effects or interaction were
found

Deceleration during lifting the foot upwards just before moving the foot downwards.

Source df F Sig.

sound Sphericity Assumed 2 .013 .987

Greenhouse-Geisser 1.656 .013 .975

Huynh-Feldt 1.803 .013 .981

Lower-bound 1.000 .013 .909

Error(sound) Sphericity Assumed 36
Greenhouse-Geisser 29.807
Huynh-Feldt 32.447
Lower-bound 18.000
repetition Sphericity Assumed 1 .734 .403
Greenhouse-Geisser 1.000 .734 .403
Huynh-Feldt 1.000 .734 .403
Lower-bound 1.000 .734 .403
Error(repetition) Sphericity Assumed 18
Greenhouse-Geisser 18.000
Huynh-Feldt 18.000
Lower-bound 18.000
sound * repetition Sphericity Assumed 2 .134 .875
Greenhouse-Geisser 1.874 .134 .862
Huynh-Feldt 2.000 .134 .875
Lower-bound 1.000 .134 .719
Error(sound*repetition) Sphericity Assumed 36

Greenhouse-Geisser 33.728

Huynh-Feldt 36.000

Lower-bound 18.000
Table 62: ANOVA – sound x repetition – no significant sound effects or interaction were
found

145
 GSR

HF2_max_
min_walk_
NF2_max_min LF2_max_min_ HF2_max_min NF2_max_min LF2_max_min z-
_march_z - march_z - _march_z - _walk_z - _walk_z - HF1_max_
NF1_max_min LF1_max_min_ HF1_max_min NF1_max_min LF1_max_min min_walk_
_march_z march_z _march_z _walk_z _walk_z z
b c c b c
Z -1.569 -.604 -1.569 -1.730 -.885 -2.575c
Asymp. Sig. (2-
.117 .546 .117 .084 .376 .010
tailed)
Table 63: Wilcoxon test for all variables to identify significant effects of sound repetition part
1

NF2_avg_m LF2_avg_mar HF2_avg_m NF2_avg_ LF2_avg_w HF2_avg_

arch_z - ch_z - arch_z - walk_z - alk_z - walk_z -
NF1_avg_m LF1_avg_mar HF1_avg_m NF1_avg_ LF1_avg_w HF1_avg_
arch_z ch_z arch_z walk_z alk_z walk_z
b b c b b
Z -1.006 -1.328 -1.690 -.724 -1.087 -1.771c
Asymp. Sig.
.314 .184 .091 .469 .277 .077
(2-tailed)
Table 64: Wilcoxon test for all variables to identify significant effects of sound repetition part
2

Difference between maximum and minimum arousal during marching

N 19
Chi-Square 9.579
df 2
Asymp. Sig. .008
Table 65: Friedman’s ANOVA

LF_max_min_ma HF_max_min_ma HF_max_min_ma

rch_z_mean - rch_z_mean - rch_z_mean -
NF_max_min_ma NF_max_min_ma LF_max_min_ma
rch_z_mean rch_z_mean rch_z_mean

Z -.885b -2.093b -2.133b

Asymp. Sig. (2-tailed) .376 .036 .033
Table 66: Wilcoxon test within the three different sound conditions

Average arousal during march

146
N 19
Chi-Square 5.158
df 2
Asymp. Sig. .076
Table 67: Friedman’s ANOVA

Average arousal during walk

N 19
Chi-Square 5.158
df 2
Asymp. Sig. .076
Table 68: Friedman’s ANOVA

Difference between maximum and minimum arousal during walking

N 20
Chi-Square .300
df 2
Asymp. Sig. .861
Table 69: Friedman’s ANOVA

147
APPENDIX H – INFORMATION SHEET

Information Sheet for Participants in Research Studies

You will be given a copy of this information sheet.

Title of Project: The hearing body

This study has been approved by the UCL Research

Ethics Committee as Project ID Number: Staff/1213/003

Name, Address and Contact Details of Maria Basia

Investigators:
[email protected], 07707471469

Dr Ana Tajadura (supervisor), [email protected],

07772026812

UCL Interaction Centre

MPEB 8th floor, University College London Gower Street,

London, WC1E 6BT

We would like to invite you to participate in this research project. You should only participate if you want to;
choosing not to take part will not disadvantage you in any way. Before you decide whether you want to take
part, please read the following information carefully and discuss it with others if you wish. Ask us if there is
anything that is not clear or you would like more information.

The study aims to explore how sensory simulation during walking affects behaviour and physiological
change. This research takes place in three parts. First you will be provided with a questionnaire for collecting
demographic data and information related to the study. Thereafter, you will be asked to perform a task six
different times, each lasting about 5 minutes. For this task you will wear headphones, through which you will
listen to environmental sounds, and a pair of sandals. We will also ask you to carry a small bag with you with
some equipment inside. In the beginning of each different task session you will be given a spanner. Then you
will be asked to march in place for 20 seconds until the experimenter gives you a “go” signal, after which you
will be required to walk down a corridor and put the spanner in a box placed at the end of the corridor.

148
Please focus on this walking task. Between each of the different six periods you will be asked to complete 3
small tasks including you to provide three different dimension estimates and to fill in a brief questionnaire.
Finally, after the end of all six sessions you will also be required to fill in a last short questionnaire. Please
note that you may omit questions that you do not need to answer.

Your participation will be video recorded. In addition, during the experiment you will be asked to wear a
number of sensors. Accelerometer and electromyography sensors, which measures your movements and
will be placed on your legs and on the upper part of your back and chest; a wrist sensor, which measures the
amount of sweat produced at your wrist (EDA – electrodermal activity) and pressure sensors attached to a
pair of sandals.

The procedure is not harmful or painful in any way. This experiment does not involve any risk for you. The
whole session lasts about 1 hour. You will be fully debriefed. You will be paid £7.50 for taking part.

All data will be collected and stored in accordance with the Data Protection Act 1998 and will be kept
anonymous. Researchers working with me will analyze the data collected.

It is up to you to decide whether or not to take part. If you choose not to participate, you won't incur any
penalties or lose any benefits to which you might have been entitled. However, if you do decide to take part,
you will be given this information sheet to keep and asked to sign a consent form. Even after agreeing to
take part, you can still withdraw at any time and without giving a reason.

149