sensors

Article
A Wearable Multi-Modal Digital Upper Limb Assessment
System for Automatic Musculoskeletal Risk Evaluation
Abdullah Tahir 1,2 , Shaoping Bai 3, * and Ming Shen 1

1 Department of Electronic Systems, Aalborg University, 9220 Aalborg, Denmark; [email protected] (A.T.);
[email protected] (M.S.)
2 Department of Mechanical, Mechatronics, and Manufacturing Engineering, University of Engineering &
Technology Lahore, Faisalabad Campus, Faisalabad 38000, Pakistan
3 Department of Materials and Production, Aalborg University, 9220 Aalborg, Denmark
* Correspondence: [email protected]

Abstract: Continuous ergonomic risk assessment of the human body is critical to avoid various
musculoskeletal disorders (MSDs) for people involved in physical jobs. This paper presents a digital
upper limb assessment (DULA) system that automatically performs rapid upper limb assessment
(RULA) in real-time for the timely intervention and prevention of MSDs. While existing approaches
require human resources for computing the RULA score, which is highly subjective and untimely,
the proposed DULA achieves automatic and objective assessment of musculoskeletal risks using
a wireless sensor band embedded with multi-modal sensors. The system continuously tracks and
records upper limb movements and muscle activation levels and automatically generates muscu-
loskeletal risk levels. Moreover, it stores the data in a cloud database for in-depth analysis by a
healthcare expert. Limb movements and muscle fatigue levels can also be visually seen using any
tablet/computer in real-time. In the paper, algorithms of robust limb motion detection are developed,
and an explanation of the system is provided along with the presentation of preliminary results,
which validate the effectiveness of the new technology.

Keywords: digital upper limb assessment (DULA); rapid upper limb assessment (RULA);
musculoskeletal disorder (MSD); MSD prevention and intervention; multi-modality wearable sensors

Citation: Tahir, A.; Bai, S.; Shen, M. A

Wearable Multi-Modal Digital Upper
Limb Assessment System for 1. Introduction
Automatic Musculoskeletal Risk Physical workload assessment is critical but complex and multi-dimensional, which
Evaluation. Sensors 2023, 23, 4863. requires careful consideration of various physical workload modalities, e.g., intensity,
https://doi.org/10.3390/s23104863 duration, frequency, and variability, together with environmental and individual factors,
Academic Editor: Ki H. Chon e.g., temperature, lighting conditions, age, sex, and physical fitness [1]. Workers in factories,
individuals performing activities of daily life (ADL) with poor posture, and even computer
Received: 21 March 2023 and mobile phone users are prone to develop MSDs [2]. MSDs include temporary or
Revised: 12 May 2023
permanent injuries in the inner body parts, e.g., muscles, tendons, joints, and spinal disks.
Accepted: 16 May 2023
Industrial workers often have to perform repetitive movements, heavy lifting, and body
Published: 18 May 2023
twisting, which can lead to a variety of problems, e.g., chronic pain, decreased mobility,
and MSDs. Such jobs require workers to work in conditions that are not conducive to good
posture and body alignment [3], hence carrying high risks of developing MSDs. A recent
Copyright: © 2023 by the authors.
survey for working conditions in Europe indicates that 42% of European workers complain
Licensee MDPI, Basel, Switzerland. of pain around the neck and upper limbs [4].
This article is an open access article To ensure the safety of workers, it is important to assess the ergonomic risks associated
distributed under the terms and with these tasks and implement controls and interventions to improve working conditions.
conditions of the Creative Commons Several risk assessment methods have been developed, such as rapid entire body assess-
Attribution (CC BY) license (https:// ment (REBA) [5], novel ergonomic postural assessment (NERPA) [6], strain index (SI) [7],
creativecommons.org/licenses/by/ Ovako working posture analyzing system (OWAS) [8], and RULA [9]. Of them, RULA is
4.0/). the most widely used subjective tool to quickly assess the ergonomic risk factors associated

Sensors 2023, 23, 4863. https://doi.org/10.3390/s23104863 https://www.mdpi.com/journal/sensors

Sensors 2023, 23, 4863 2 of 14

with upper limbs, incorporating both muscle utilization and limb postures. Basahel et al.
used RULA to assess ergonomic risk factors associated with lifting and pulling tasks in an
industrial environment [10]. In another study, a modified RULA method was proposed for
two posture risk quantification methods, namely event-based and time-based, for upper-
extremity MSDs [11]. Additional similar works using RULA and its modified forms to
assess ergonomic risk factors associated with industrial workers can be found in refer-
ences [12,13]. Since RULA requires deployment in the field, therefore these non-automatic
methods require tedious setup and a medical expert to continuously observe workers’ body
movements. To empower devices to take on the burden from humans, body-mounted
sensors can be used for computing RULA scores automatically, e.g., electromyograph
(EMG) based methods [14–16] and inertial-sensor-based methods [17–19]. However, it is
essential to note that EMG-based methods have higher costs and longer deployment times.
The past works utilizing inertial-based methods do not explicitly provide kinematic details
and use graphic software, e.g., CATIA or AnyBody, to show limb movements, which put a
limitation on the design and deployment of an automatic RULA assessment system in the
form of an embedded wearable device.
As a graphical assessment technique, a digital human model (DHM) can be used
to visualize and analyze the movements of limbs and human posture in computer soft-
ware [20,21]. Such techniques are particularly useful to identify and mitigate potential risks
before designing a product or workstation. This technology is suitable for early ergonomic
analysis but cannot be utilized in real-time, as performing a certain task is majorly depen-
dent upon an individual’s preference. Therefore, the utilization of muscles or body posture
would be different for the same task from time to time and person to person. Hu et al.
worked on predicting the real-world ergonomic measurements by simulating the scenario
in a virtual environment (VE) while a subject did the movements in a real environment [22].
Nevertheless, VE-based methods can only be deployed and tested in a laboratory setup.
Another mode of risk assessment is based on video recording [23–25]. Optical assessment
methods are tedious to set up with the constraint of having direct sight of the worker. Other
assessment methods, e.g., questionnaires and interviews [26,27], limit their practicality due
to the substantial need for human resources, and such methods can be biased.
For the people working in environments where the setup is already installed, non-
obstructive and intrinsic or body-mounted sensors, e.g., EMG, inertial sensors, gyroscope,
and gravimeters, can be used for recording and analyzing motion data, and further for
ergonomic risk assessments [17–19,28]. Padilla et al. designed a web-based risk-factor
assessment system using inertial sensors and displayed limb movements [29], but they
did not provide any details for limb kinematics. Employing graphic software to display
limb movements is common but not very helpful in assessing working conditions due to
the lack of kinematic information or evaluating exoskeleton design, particularly passive
exoskeletons [16,30] in industrial usages. A comparison of past works is listed in Table 1.

Table 1. List of related works.

Load
Reference Assessment Tool Data Acquisition Method Wearable Non-Obstructive
Identification
RULA/Modified
[10–13,23,24] Optical/Self-report No No No
RULA/SI
[14–16] Muscle activity level EMG/AnyBody software Yes Yes Yes
[20–22] RULA DHM/VR/CATIA software Yes No No
[17,19] RULA/REBA/NIOSH Inertial Yes Yes No
Artificial Intelligence
[18] Inertial and optical Yes No No
based
Our work RULA Inertial Yes Yes Yes
Sensors 2023, 23, 4863 3 of 14

In this study, a multi-modal digital upper limb assessment system—DULA—is pre-

sented along with robust kinematics, which continuously tracks and records muscle ac-
tivities and limb motions to digitally fill the RULA chart and automatically generate risk
levels. DULA consists of multi-modal sensor bands having wireless sensors interfaced
with a mobile or tablet application. Two modalities, (i) limb posture in space using IMUs,
and (ii) muscle activity through a force myograph (FMG), are devised for the proposed
system. A robust kinematic model is developed to accurately determine upper limb motion.
Furthermore, the data from these wearable sensors is continuously recorded and stored in
Google Firebase, which can be retrieved for visualization, examination, and further analysis
by a medical expert. The contributions of this work are:
1. Multi-modal wireless wearable sensors are used for digitally computing RULA scores.
2. A robust kinematic model has been developed for upper limbs motion tracking.
3. Digital implementation of the new method has been validated by comparison with an
online RULA calculator.
The balance of the paper is organized as follows: Section 2 describes RULA, kinematics
to detect arm posture, frame transformations, and load identification. Hardware and
software implementation is presented in Section 3. Results are presented in Section 4 and
discussed in Section 5. Finally, conclusions are drawn in Section 6.

2. DULA Method
The DULA system utilizes battery-powered wireless multi-modal sensor bands, which
are wirelessly connected to a mobile application through Bluetooth. As will be presented in
Section 3, each sensor band is equipped with an IMU and force-sensitive resistors (FSRs)
for FMG, which are primarily used for ergonomic risks analysis, whereas gyroscopes
and gravimeters are used to set up coordinate frames with the upper limbs for kinematic
analysis. The system automatically generates the RULA score, which will be henceforth
referred to as the DULA score.

2.1. RULA
DULA is a digital implementation of RULA. We will, therefore, briefly describe the
RULA method for completeness’ sake. RULA is a subjective observation method of posture
analysis to assess the risks of developing MSDs for individuals who perform physical
tasks using their upper extremities. Illustrations for different upper limb postures, the load
being lifted, and associated numerical scores are shown in Figure 1 [9]. The DULA system
finds the orientation of the upper arm and forearm and the load being lifted, assigns the
corresponding scores, and yields the aggregated score. The aggregated score has the range
of 1–9, which determines the overall risk of MSDs for upper limbs [9] and the corresponding
action level as shown in Table 2. A lower RULA score corresponds to a lower risk of
developing MSDs indicating that work posture is good enough. Hence, no change in the
working scheme is required. In contrast, a higher RULA score corresponds to a severe risk
of developing MSDs, and therefore, the working scheme should be immediately changed.

Table 2. RULA score, action level, and corresponding action.

RULA Score Action Level Action

1–2 1 Acceptable working pattern; no changes are required.
3–4 2 Working pattern may be changed; hence, further investigations
are suggested.
5–6 3 Working pattern will soon require changes; therefore, further
investigations are needed.
>7 4 Working pattern highly risks health and should be immediately
changed.
Sensors 2023, 23, 4863 4 of 14

Upper arm angle "! (deg) Score Load / (kg) Score

"! < −20 2 /<2 0
−20 ≤ "! < 20 1 2 < / < 10 (intermittent) 1
%# 20 ≤ "! < 45 2 2 < / < 10 (static or repeated) 2
45 ≤ "! < 90 3 / > 10 or repeated or shocks 3
"! > 90 4
(a) (c)

Forearm angle "# (deg) Score

0 < "# < 60 2
60 ≤ "# < 100 1

%! "# ≥ 100 2

(b) (d)

Figure 1. DULA diagram for ergonomic assessment. (a) Upper arm postures and corresponding
scores, (b) forearm postures and corresponding scores, (c) load being lifted and corresponding scores,
(d) wearable sensor band.

2.2. Kinematics for DULA

DULA, i.e., automatic computation of RULA score, requires first determining the
accurate forearm and upper arm orientation. IMUs in the sensor band are primarily used
for this purpose. However, raw data from IMUs do not provide any useful kinematic
information without the attachment of reference coordinate frames on the upper arm and
forearm; doing so will provide a coordinate reference to measure the limb’s orientation.
Setting up the coordinate frame is explained in Section 2.2.1 while frame transformations
of the upper arm and forearm movements with respect to the attached coordinate frames
are explained in Sections 2.2.2 and 2.2.4.

2.2.1. Reference Coordinate Setup

The reference coordinate frames for the forearm and upper arm are shown in Figure 2.
To attach a frame to the forearm, a sensor band is worn on the dorsal side of the arm,
i.e., near the wrist. A certain posture and a certain movement are performed so that a
coordinated setup can be established. Each course of action identifies one unit vector of
the forearm as each movement projects one unit vector of the forearm reference frame in
the coordinate frame identified by the current orientation of IMU referred to as { I }. The
cross-multiplication of two orthogonal unit vectors from the two actions yields a third unit
vector or axis and completes the procedure. The right forearm frame is settled with the
following steps:
1. Stretch the arm forward while the palm is facing downwards. This is considered the
right arm base posture.
2. Record the gravity vector g at base posture, upon which the z Fˆ ∗ for the forearm frame
is determined.
g
z Fˆ ∗ = − . (1)
k gk
3. Perform supination motion and record angular velocity ω during the motion. The y ˆF∗
for the forearm frame is hence obtained as:
ω
y ˆF∗ = − . (2)
kω k
Sensors 2023, 23, 4863 5 of 14

4. The cross product of y ˆF∗ and z Fˆ ∗ results in x F∗ for the forearm frame. However, it is
not humanly possible to have y ˆF∗ and z Fˆ ∗ exactly perpendicular to each other, hence,
x F∗ is normalized.
x F∗ = y ˆF∗ × z Fˆ ∗ , (3)

x F∗
xˆF∗ = . (4)
k x F∗ k
5. Finally, z Fˆ ∗ is obtained to complete right hand coordinate frame convention.

z Fˆ ∗ = xˆF∗ × y ˆF∗ . (5)

6. Thus, the computed rotation matrix belongs to a special orthogonal group SO(3). The
forearm reference frame in the IMU frame has the following form:
F∗
 
I R = xˆF∗ y ˆF∗ z Fˆ ∗ . (6)

The method to attach a reference coordinate frame on the right upper arm follows
the same steps as the forearm, except that the sensor band is worn on the biceps muscles.
The same steps are repeated for the left forearm and left upper arm; however, for the left
limbs, pronation motion will be performed as in Step 3, and the angular velocity direction
is taken positively.

$!
!∗
$!
#∗
{F*}
{U*}

#!
!∗ #!
#∗
"!
!∗ "!
#∗

Forearm band
(dorsal side of forearm)

Upper arm band

(over biceps)

Figure 2. Reference coordinate frames setup for the forearm { F ∗ } and the upper arm {U ∗ }.

2.2.2. Frame Transformations

A robust frame transformation methodology is developed, which provides the limbs
orientations with respect to corresponding reference coordinate frames. A scenario, as
shown in Figure 3, is presented for the right forearm, where the frames, shown in curly
braces, are randomly oriented to formulate the mathematical expression of the rotation
matrix to represent the orientation of the forearm { F } with respect to the forearm reference
frame { F ∗ }. The matrix R represents a rotational transformation from the frame in the
subscript to the frame in superscript, whereas the subscripts under the axes identify the
frame they belong to.
Sensors 2023, 23, 4863 6 of 14

)
(1 YG
∗
{I} *
)1
Z* F, ZI
{G}
{F*} XG
X* F
ZG
Y* F YI *
(1
XI

{F} Right hand posture during coordinate frame

ZF setup. IMU frame {I} and forearm frame {F*}
are attached.

A random right hand orientation after

coordinate frame setup.
XF

YF Into the plane

Out of the plane

Figure 3. Frame setup for the kinematics of arm motion. The coordinate frames are shown in
curly braces, whereas rotation matrices adjacent to arrows connecting two frames represent their
transformation. { G } is the global reference frame. { F ∗ } and { I } are the forearm reference frame and
IMU frame, respectively, and they are attached to the initial configuration of the forearm. Frame { F }
moves with the forearm.

Consider the posture when the sensor band is worn on the forearm. Raw data from
the IMU represents the forearm orientation in global reference frame G F R. A reference frame
∗
{ F } is attached to the forearm according to the method described in Section 2.2.1. The
objective is to find FF∗ R, however, it should be noted that { F ∗ } does not necessarily exhibit
the same orientation as that of { G }. Furthermore, the reference frame { F ∗ } constitutes a
new coordinate system. Hence, the objective is to find the forearm frame { F } orientation
which is defined with respect to { G }, in a new coordinate system established by { F ∗ }. The
method is described as follows:
1. Record IMU orientation G I R when the arm is in the base posture.

2. F ∗ R is computed according the method in Section 2.2.1.

I
3. F
G R is continuously acquired from the IMU sensor in real-time.
4. To find the forearm orientation with respect to IMU frame, i.e, FI R, the following
equations are presented:
F I F
G R = G R I R, (7)

 T
F I F
IR = GR G R. (8)

5. Finally, we find the orientation of the forearm with respect to the forearm reference
frame { F ∗ },
 ∗ T
F F F F∗
F∗ R = I R I R I R. (9)

The same methodology is applied to find upper arm transformation to have a rotation
matrix describing upper arm {U } orientation with respect to upper arm reference frame
{U ∗ } .
Sensors 2023, 23, 4863 7 of 14

2.2.3. Upper Arm Orientation

The rotation matrix defining upper arm orientation with respect to the upper arm
reference frame is considered to be formed by the upper arm extension/flexion ψu , abduc-
tion/adduction θu , and internal/external rotation ψu . Mathematically:
 
r11 r12 r13
U
U∗ R = R XZ0 Y00 = R X (φu ) R Z0 (θu ) RY00 (ψu ) = r21 r22 r23 , (10)
r31 r32 r33

where, starting with a frame coincident with upper arm reference frame {U ∗ }, R X , R Z0 (θu ),
and RY 00 (ψu ) represent three consecutive rotations following XZY convention. The angles
φu , θu , and ψu are obtained as:

φu = arctan (r32 , r22 ), (11)

θu = arcsin (−r12 ), (12)

ψu = arctan (r13 , r11 ). (13)

2.2.4. Forearm Orientation

To compute the RULA score for a forearm, the relative rotation of the forearm with
respect to upper arm is required. However, the orientations of the forearm and upper
arm are defined in { F ∗ } and {U ∗ }, respectively, while both constitute different coordinate
systems. Therefore, we map the forearm orientation in the coordinate frame {U ∗ } and then
find the relative rotation.
1. Find the orientations of upper arm base frame {U ∗ } and forearm base frame { F ∗ }
with respect to global reference frame { G } as:

F∗ If F∗
G R = G R I f R, (14)

U∗ Iu U ∗
G R = G R Iu R, (15)
where Iu and I f are upper arm IMU and forearm IMU frames, respectively.
2. Compute the relative rotation R1 between both frames { F ∗ } and { G ∗ }:

F∗ U∗
G R R1 = G R, (16)

T
F∗ U∗

R1 = G R G R. (17)

3. The forearm orientation { F } defined in { F ∗ } is mapped in the upper arm reference

frame {U ∗ } using similarity transformation technique:
F
U∗ R = R1T F
F∗ R R1 . (18)

4. Since both forearm and upper arm orientations are now established in the same
coordinate system defined by {U ∗ }, the relative rotation matrix R2 is found as:
U F
U∗ R R 2 =U ∗ R, (19)

 T
U F
R2 = U∗ R U ∗ R. (20)
Sensors 2023, 23, 4863 8 of 14

5. The forearm orientation with respect to the upper arm is considered to be formed by
the combination of two rotations, the forearm extension/flexion α f about the current
Y, and the forearm pronation/supination β f about the current Z axis.
 
s11 s12 s13
R2 = RYZ0 = RY (α f ) R Z0 ( β f ) = s21 s22 s23 . (21)
s31 s32 s33

The composition of orientation is illustrated in Figure 4. Inverse kinematics is used to

find the forearm angles:
α f = arctan (s21 , s22 ), (22)

β f = arctan (s13 , s33 ). (23)

ZU
0,
YU
XU
XF ZF
ZU '

XU ' 1,
YU '
YF

Figure 4. Moving from the upper arm frame to the forearm frame, when forearm orientation is
already defined with respect to frame {U ∗ }.

2.3. Load Identification

FMG is a technology that can measure changes in muscle volume during limb motion,
which arise from muscle contractions and expansions. An FSR is a special resistor that
exhibits a change in resistance when a force is applied to it. Therefore, FMG is primarily
done using FSRs, and each sensor band is equipped with eight FSRs. Upper limb muscles
hold the weight carried by an individual since the sensor band is worn over bicep muscles;
hence the change in measured resistance corresponds to the muscle activity, which is
interpreted as force being exerted or load being lifted by an individual. Various research
works are available to identify payload using FMG [31,32]. When the sensor band is worn
over the upper arm, the maximum value for each FSR channel is recorded by intense elbow
flexion, later used to normalize FSR data. Normalized FSR values for 2 kg and 10 kg
loads, as two thresholds for RULA, are important according to Figure 1c and are marked
experimentally to identify the load category. Finally, the identified load is scored according
to Figure 1c.

3. DULA System Development

A portable and easy-to-deploy multi-modal digital upper limb assessment system,
DULA, has been developed for automatic musculoskeletal risk assessments as shown in
Figure 5. DULA continuously tracks and records upper limbs motion as well as upper arm
Sensors 2023, 23, 4863 9 of 14

muscle activity. The system generates single-digit risk assessment scores, ranging from 1 to
9, which indicates the risk of developing MSDs. A lower score means that human postures
for carrying out physical tasks are appropriate, implying that no change in the working
pattern is required, while a higher score corresponds to higher chances of developing MSDs
implying that the working pattern should be immediately changed. The system also stores
human motion data and muscle activity data in Google Firebase for visualization and
in-depth analysis by medical experts.

Forearm sensor Upper arm sensor

CONNECTED CONNECTED

Upper arm Forearm Weight

43.2, 21.6, 13.7 34.2, 13.2 > 2 kg

Score
3

Signal 9.0
0.0
conditioning ADC
FSR IOs

I2C
9-axis IMU

I2C
ESP32
3-axis gravity

DAQ in arm band.

Figure 5. System hardware and software overview.

3.1. Hardware
The hardware of the designed system consists of wireless, battery-powered, multi-
modal sensor bands and a mobile phone or tablet. The bands are connected to mobile/tablet
applications through Bluetooth. Each sensor band has various sensors, e.g., an IMU
consisting of a gravity sensor, gyroscope, linear and angular accelerometers, and eight
FSRs. All the sensors provide kinematic information except FSRs, which are used for FMG
to determine the weight being lifted by an individual. FSRs are embedded in flexible
armbands to have a tight grip over muscles. Data from FSRs are amplified and passed
through low pass filters before feeding to analog to onboard digital conversion (ADC)
modules, while data from IMU are acquired using the I2C protocol of the ESP32 chip.
This chip then sends the data to the mobile phone in real-time through the onboard
Bluetooth module.
Sensors 2023, 23, 4863 10 of 14

3.2. Software
A mobile application is developed to receive data from sensor bands and to perform
the matrix operations for digital upper limb assessment in real-time. On startup, this
application guides the user through the procedure of attaching the reference coordinate
frames with left and right forearms and upper arms. Once coordinate setup is complete,
multi-modal data is acquired, and matrix operations are performed in the background
of the mobile application. Risk assessments, limb movements, and muscle usage are
graphically shown on the screen. The application has an additional feature of triggering an
alarm/notification if an individual is at a high risk of developing MSDs. The application is
developed using Flutter and vector_math_64 library is used for matrix operations.

4. Results
With the developed algorithm and system, a DULA score has been generated for
the right upper limbs as a test scenario where a person picked an electrical screwdriver
from a table and tightened a screw on the wall. Figure 6 shows the body posture for the
screw-tightening task and corresponding limb orientations. Angular velocity and gravity
data are used only for coordinate frames setup as described in Section 2.2.1 and are not
shown here. The orientation of the upper arm and forearm, upper arm abduction, forearm
working out to the side of the body or across body mid-line, wrist twist, and carried weight
are the factors considered to digitally fill up the RULA score chart. The DULA score is
generated in accordance with Figure 1, leading to the generation of the final RULA score
according to the method described in Reference [9]. Figure 6 shows that the designed
experiment consists of five major movements. The DULA score for each phase shows that
there are no ergonomic risks associated with standing still (SS), while approach phases (A1
and A2) show DULA scores of 3 and 4, respectively, indicating that these postures and
movements do not carry high risks for developing MSDs.
Correlating the DULA score for the screw tightening task with Table 2 indicates that
the working routine will soon require changes, and the long time exposure in carrying out
this task in the same pattern will certainly lead to developing MSDs. Hence investigations
to improve the task procedure or workstation are suggested. To validate the proposed
methodology, a web-based ergonomic assessment tool for RULA [33] is used to manually
input upper limb postures to generate corresponding RULA scores. Ten different body
postures as shown in Figure 7, and three different weights (1.5 kg, 6 kg, and 12 kg) are set
for an individual, and RULA scores from the web-based tool and from the DULA system
are compared as shown in Table 3. The verification of DULA is ensured by matching the
generated scores, as the difference in the RULA scores was 0.3 ± 0.49.

Table 3. RULA and DULA scores.

Posture DULA RULA Posture DULA RULA

1 6 6 6 6 6
2 5 5 7 7 6
3 5 4 8 5 5
4 4 4 9 6 6
5 6 5 10 7 7
Sensors 2023, 23, 4863 11 of 14

Lifted load

Load (kg)
Time (s)
Upper arm orientation

Angle (deg)

Time (s)

Screw tightening task Forearm arm orientation

Angle (deg)

Time (s)
RULA score
Magnitude

SS: Stand still

A1: Approach screw site Time (s)
ST: Screw tightening
A2: Approach normal posture
SS A1 ST A2 SS

Figure 6. Screw tightening task, corresponding data from the sensors, and RULA score.

1 2 3 4 5

6 7 8 9 10

Figure 7. Ten scores for ten different postures and lifting three different weights.
Sensors 2023, 23, 4863 12 of 14

5. Discussion
RULA is well developed ergonomic assessment tool, effective for MSD prevention. In
this study, a new system, named DULA, is designed for digital rapid upper limb assessment
for MSD early prevention and intervention. The system can be used by physical workers
and can be easily deployed in any working setup. In addition, the system has the potential
to be used with assistive devices, e.g., wearable exoskeletons [30], for their performance
assessment, ensuring their effective use. For ergonomic risk assessments in industrial
environments, different studies have evaluated ergonomic risks by recording workers’
movements, through questionnaires, or manual analysis by medical experts, which are
time-consuming and delicate evaluations. The DULA system effectively addresses these
challenges using wearable sensors linked with a mobile application and performing the
risk assessment in real-time. However, RULA may not be suitable for assessing ergonomic
risks associated with very complex or specialized tasks, which may require a more tailored
approach to indicate a correct risk evaluation and would also require more wearable sensor
bands. Nevertheless, RULA combined with fuzzy logic has the potential to indicate a better
risk level. The same methodology can be further enhanced by tracking and scoring lower
limbs, wrists, and necks.
It has to be noted that the risk of developing MSDs is also influenced by various body-
related factors, such as age, height, weight, and any comorbidities that the worker may
have. RULA provides quick and valuable insights into the potential for injury or discomfort
associated with a particular job or task. However, it has limitations in considering the time
for performing the task and environmental conditions, e.g., temperature, lighting, noise, etc.
Upper arm muscles are responsible for lifting weight. Their role is particularly significant
when there is elbow flexion. However, when the elbow is not flexed, then instead of upper
arm muscles, weight is carried by bones or by shoulder muscles. In such circumstances,
FSRs over bicep muscles are not helpful in identifying the right value of the picked load.
Regarding the measurement of muscle activation to determine the lifted load, limitations
occur when FSRs become saturated or FSRs experience external force. Consequently, FMG
data becomes erroneous. Anti-saturation is required for the post-processing of FMG data.
Another limitation is related to an inherent singularity with Euler angles. Euler angles have
the intrinsic property that they undergo singularity when the rotation about the middle
axis makes the first and third axes parallel to each other.

6. Conclusions
The proposed system, DULA, digitally records and automatically analyzes upper
limb motions and muscle activities to determine the RULA score for the upper limbs
without any intervention from a medical expert. Preliminary results have shown that
DULA comprehends RULA. For the ten different postures and three different weights
(1.5 kg, 6 kg, and 12 kg), the difference between RULA and DULA is 0.3 ± 0.49.
DULA is a portable system that can be easily deployed in industry and possesses great
potential to protect workers from developing temporary or permanent MSDs through early
intervention and prevention. In future research, the following issues will be considered
to further enhance the methodology: (a) exploring the impact of external forces on FMG,
(b) incorporating the analysis of wrist, neck, and lower limbs to enable comprehensive
and automated whole-body posture analysis, and (c) integrating task duration into RULA
to enhance the accuracy of the assessment. These research directions hold great potential
in advancing the field of workplace ergonomics and contributing to the development of
effective interventions to prevent musculoskeletal disorders.

Author Contributions: Conceptualization, S.B.; methodology, A.T. and S.B.; validation, A.T.; formal
analysis, A.T. and S.B.; investigation, A.T. and S.B.; resources, S.B.; writing—original draft preparation,
A.T.; writing—review and editing, S.B. and M.S.; visualization, A.T., S.B. and M.S.; supervision, S.B.
and M.S. All authors have read and agreed to the published version of the manuscript.
Funding: This research received funding from Interreg North Sea Region.
Sensors 2023, 23, 4863 13 of 14

Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Not applicable.
Acknowledgments: The authors would like to extend their sincere gratitude to the Interreg North Sea
Region-funded project EXSKALLERATE. The authors thank Pascal Madeleine from the Department
of Health Science and Technology, Aalborg University, for his discussion of and comments on the
paper and Mathias Mosskov Jacobsen from the Department of Computer Science, Aalborg University,
for developing the mobile application.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:

MSD Musculoskeletal disorders

RULA Rapid upper limb assessment
DULA Digital upper limb assessment
IMU Inertial measurement unit
F or f Forearm
U or f Upper arm

References
1. Winkel, J.; Mathiassen, S.E. Assessment of physical work load in epidemiologic studies: Concepts, issues and operational
considerations. Ergonomics 1994, 37, 979–988. [CrossRef]
2. Dockrell, S.; O’Grady, E.; Bennett, K.; Mullarkey, C.; Mc Connell, R.; Ruddy, R.; Twomey, S.; Flannery, C. An investigation of the
reliability of Rapid Upper Limb Assessment (RULA) as a method of assessment of children’s computing posture. Appl. Ergon.
2012, 43, 632–636. [CrossRef] [PubMed]
3. Lei, L.; Dempsey, P.G.; Xu, J.G.; Ge, L.N.; Liang, Y.X. Risk factors for the prevalence of musculoskeletal disorders among Chinese
foundry workers. Int. J. Ind. Ergon. 2005, 35, 197–204. [CrossRef]
4. Parent-Thirion, A.; Biletta, I.; Cabrita, J.; Vargas Llave, O.; Vermeylen, G.; Wilczyńska, A.; Wilkens, M. Sixth European Working
Conditions Survey–Overview report. Eurofound 2016. [CrossRef]
5. Hignett, S.; McAtamney, L. Rapid entire body assessment (REBA). Appl. Ergon. 2000, 31, 201–205. [CrossRef]
6. Sanchez-Lite, A.; Garcia, M.; Domingo, R.; Angel Sebastian, M. Novel ergonomic postural assessment method (NERPA) using
product-process computer aided engineering for ergonomic workplace design. PLoS ONE 2013, 8, e72703. [CrossRef]
7. Moore, J.S.; Vos, G.A. The strain index. In Handbook of Human Factors and Ergonomics Methods; CRC Press: Boca Raton, FL, USA,
2004; pp. 109–114.
8. Karhu, O.; Kansi, P.; Kuorinka, I. Correcting working postures in industry: A practical method for analysis. Appl. Ergon. 1977,
8, 199–201. [CrossRef]
9. McAtamney, L.; Corlett, E.N. RULA: A survey method for the investigation of work-related upper limb disorders. Appl. Ergon.
1993, 24, 91–99. [CrossRef]
10. Basahel, A.M. Investigation of work-related musculoskeletal disorders (MSDs) in warehouse workers in Saudi Arabia. Procedia
Manuf. 2015, 3, 4643–4649. [CrossRef]
11. Bao, S.; Howard, N.; Spielholz, P.; Silverstein, B. Two posture analysis approaches and their application in a modified rapid upper
limb assessment evaluation. Ergonomics 2007, 50, 2118–2136. [CrossRef]
12. Intranuovo, G.; De Maria, L.; Facchini, F.; Giustiniano, A.; Caputi, A.; Birtolo, F.; Vimercati, L. Risk assessment of upper limbs
repetitive movements in a fish industry. BMC Res. Notes 2019, 12, 1–7. [CrossRef] [PubMed]
13. Valentim, D.P.; de Oliveira Sato, T.; Comper, M.L.C.; da Silva, A.M.; Boas, C.V.; Padula, R.S. Reliability, construct validity and
interpretability of the Brazilian version of the Rapid Upper Limb Assessment (RULA) and Strain Index (SI). Braz. J. Phys. Ther.
2018, 22, 198–204. [CrossRef] [PubMed]
14. Balasubramanian, V.; Swami Prasad, G. An EMG-based ergonomic evaluation of manual bar bending. Int. J. Ind. Syst. Eng. 2007,
2, 299–310. [CrossRef]
15. González, A.; Barrios-Muriel, J.; Romero-Sánchez, F.; Salgado, D.; Alonso, F. Ergonomic assessment of a new hand tool design for
laparoscopic surgery based on surgeons’ muscular activity. Appl. Ergon. 2020, 88, 103161. [CrossRef]
16. Kong, Y.K.; Choi, K.H.; Cho, M.U.; Kim, S.Y.; Kim, M.J.; Shim, J.W.; Park, S.S.; Kim, K.R.; Seo, M.T.; Chae, H.S.; et al. Ergonomic
assessment of a lower-limb exoskeleton through electromyography and Anybody modeling system. Int. J. Environ. Res. Public
Health 2022, 19, 8088. [CrossRef]
17. Huang, C.; Kim, W.; Zhang, Y.; Xiong, S. Development and validation of a wearable inertial sensors-based automated system for
assessing work-related musculoskeletal disorders in the workspace. Int. J. Environ. Res. Public Health 2020, 17, 6050. [CrossRef]
Sensors 2023, 23, 4863 14 of 14

18. Moore, J.; Stuart, S.; McMeekin, P.; Walker, R.; Celik, Y.; Pointon, M.; Godfrey, A. Enhancing free-living fall risk assessment:
Contextualizing mobility based IMU data. Sensors 2023, 23, 891. [CrossRef]
19. Akhmad, S.; Arendra, A.; Findiastuti, W.; Lumintu, I.; Pramudita, Y.D. Wearable IMU Wireless Sensors Network for Smart
Instrument of Ergonomic Risk Assessment. In Proceedings of the 6th Information Technology International Seminar (ITIS),
Surabaya, Indonesia, 14–16 October 2020; pp. 213–218.
20. Mohamad, D.; Md Deros, B.; Ismail, A.R.; Daruis, D.D.I.; Sukadarin, E.H. RULA analysis of work-related disorder among
packaging industry worker using digital human modeling (DHM). In Advanced Engineering Forum; Trans Tech Publications Ltd.:
Kuala Lumpur, Malaysia, 2013; Volume 10, pp. 9–15.
21. Hussain, M.M.; Qutubuddin, S.; Kumar, K.P.R.; Reddy, C.K. Digital human modeling in ergonomic risk assessment of working
postures using RULA. In Proceedings of the International Conference on Industrial Engineering and Operations Management,
Bangkok, Thailand, 5–7 March 2019; pp. 5–7.
22. Hu, B.; Ma, L.; Zhang, W.; Salvendy, G.; Chablat, D.; Bennis, F. Predicting real-world ergonomic measurements by simulation in a
virtual environment. Int. J. Ind. Ergon. 2011, 41, 64–71. [CrossRef]
23. Cao, W.; Zhong, J.; Cao, G.; He, Z. Physiological Function Assessment Based on Kinect V2. IEEE Access 2019, 7, 105638–105651.
[CrossRef]
24. Yu, Y.; Yang, X.; Li, H.; Luo, X.; Guo, H.; Fang, Q. Joint-level vision-based ergonomic assessment tool for construction workers. J.
Constr. Eng. Manag. 2019, 145, 04019025. [CrossRef]
25. Forsman, M.; Hansson, G.Å.; Medbo, L.; Asterland, P.; Engström, T. A method for evaluation of manual work using synchronised
video recordings and physiological measurements. Appl. Ergon. 2002, 33, 533–540. [CrossRef] [PubMed]
26. David, G.C. Ergonomic methods for assessing exposure to risk factors for work-related musculoskeletal disorders. Occup. Med.
2005, 55, 190–199. [CrossRef] [PubMed]
27. Cheng, H.Y.K.; Cheng, C.Y.; Ju, Y.Y. Work-related musculoskeletal disorders and ergonomic risk factors in early intervention
educators. Appl. Ergon. 2013, 44, 134–141. [CrossRef] [PubMed]
28. McDevitt, S.; Hernandez, H.; Hicks, J.; Lowell, R.; Bentahaikt, H.; Burch, R.; Ball, J.; Chander, H.; Freeman, C.; Taylor, C.; et al.
Wearables for biomechanical performance optimization and risk assessment in industrial and sports applications. Bioengineering
2022, 9, 33. [CrossRef]
29. Padilla, B.E.O.; Glushkova, A.; Menychtas, D.; Manitsaris, S. Designing a web-based automatic ergonomic assessment using
motion data. In Proceedings of the 12th ACM International Conference on Pervasive Technologies Related to Assistive
Environments, Rhodes, Greece, 5–7 June 2019; pp. 528–534.
30. Balser, F.; Desai, R.; Ekizoglou, A.; Bai, S. A novel passive shoulder exoskeleton designed with variable stiffness mechanism.
IEEE Robot. Autom. Lett. 2022, 7, 2748–2754. [CrossRef]
31. Islam, M.; Bai, S. Payload estimation using forcemyography sensors for control of upper-body exoskeleton in load carrying
assistance. Model. Identif. Control (Online) 2019, 40, 189–198. [CrossRef]
32. Tahir, A.; An, Z.; Bai, S.; Shen, M. Robust payload recognition based on sensor-over-muscle-independence deep learning for the
control of exoskeletons. IEEE Trans. Circuits Syst. II Express Briefs 2023. [CrossRef]
33. Ergonomics, O. RULA—Rapid Upper Limb Assessment. 2019. Available online: https://www.rula.co.uk/assessment.html
(accessed on 12 May 2023).

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.