Kinematics and dynamics analysis of lower limbs

based on human motion data

Ningning Hu Aihui Wang
School of Mechatronic Engineering and Automation School of Electric Information Engineer
Shanghai University Zhongyuan University of Technology
Shanghai, China Zhengzhou, China
[email protected] [email protected]
2020 Chinese Automation Congress (CAC) | 978-1-7281-7687-1/20/$31.00 ©2020 IEEE | DOI: 10.1109/CAC51589.2020.9327180

Abstract—For patients with an aging population and patients to exercise can not only be used for the verification of
suffering from movement disorders, the combination of robotic biomechanical models, but also can be used to analyze the
systems and biomedical engineering principles has brought good dynamics of the human body. The human body is a complex
news to the majority of patients. The analysis of human gait
laws can help better develop and research robotic systems. structure with multiple degrees of freedom of rotation joints.
Therefore, this article is based on the idea of using the gait Generally, the motion and trajectory of the human body are
trajectory of a healthy person as the input reference trajectory recorded through motion capture technology. After the data
of the robot, using NOKOV three-dimensional infrared motion is analyzed and processed, the human body’s spatial position,
capture equipment and three-dimensional force measurement speed, acceleration, angle, angular velocity and angular accel-
platform equipment to collect human gait movement data and
plantar force data, and analyze the collected data Kinematics eration can be accurately obtained[5].
and dynamics characteristics of human body data. The existing motion capture technologies are mainly divid-
Index Terms—gait trajectory, data processing, kinesiology, ed into four categories: mechanical motion capture systems,
human biomechanics acoustic motion capture systems, electromagnetic motion cap-
ture systems and optical motion capture systems. Although the
I. I NTRODUCTION mechanical motion capture system has low cost and simple
As we all know, it is a relatively safe and effective way to system calibration, due to the size and weight of mechanical
improve the motor function of patients’ lower limbs through equipment, many actions cannot be completed after the human
the nursing and auxiliary training of medical staff. However, body is worn, which is very inconvenient to use [6]. Acoustic
due to the aging of the population, the rate of increase in motion capture system research and development costs are
patients far exceeds the ability of medical staff to perform low, but the real-time performance is poor when the collected
auxiliary function training for all patients with motor dys- data is transmitted during motion capture, and the equipment
function [1]. The emergence of exoskeleton robots based on accuracy is also very low. The calibration of the electromag-
robotic systems and biomedical engineering principles has netic motion capture system device is simple, but the capture
brought good news to the majority of patients, and can help range is small, which is not suitable for strenuous sports.
patients to carry out a certain degree of effective rehabilitation The advantages of the three-dimensional(3D) optical motion
training without being cared for [2]. Through exoskeleton capture system are large capture range, unlimited, easy to use,
robot rehabilitation training, the patient’s lower limbs can and high sampling rate and accuracy [7-8]. The disadvantage is
move or walk like healthy people is the purpose of designing that this system is more expensive, sensitive to the sunlight and
lower limb rehabilitation robots. Therefore, the gait motion reflective equipment of the environment, and often requires
data of the human body can be used in the design and manual processing of the collected trajectory later.
real-time training of the human-like controller of the lower
limb rehabilitation robot after optimization processing. How B. 3D motion capture system
to effectively process the captured human gait motion data in Through the comparison of domestic motion capture sys-
the early stage and analyze the human body kinematics and tems, this paper adopts the NOKOV 3D infrared passive
biomechanics based on the data has become a current research optical motion capture system with better effect. Its system
hotspot [3-4]. components mainly consist of 6 Mars2H lenses (as shown in
Fig.1), connectors, tripods, three-way pan/tilt, connecting wire,
II. A NALYSIS OF THE MECHANISM OF HUMAN LOWER
calibration rod (T-shaped and L-shaped calibration rods), NK-
LIMB MOVEMENT
Cortex software, and reflective markers (Marker).
A. Current status of human motion capture equipment The motion capture system selected in this paper is designed
In the field of human sports rehabilitation technology, the with the principle of passive infrared light [9], and its sampling
measurement of the body movement process is one of the frequency can be adjusted in real time as required. The high-
current problems, because some detailed information related power HLED emitter of the Mars2H lens can make the light

978-1-7281-7687-1/20/$31.00 2020 IEEE 6727

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY TIRUCHIRAPALLI. Downloaded on July 25,2024 at 08:55:11 UTC from IEEE Xplore. Restrictions apply.
 Fig. 1. Mars2H optical camera. Fig. 2. 3D force measurement platform.

source more concentrated, efficient and durable. The image and 3 meters in height, the cameras are placed in 3 equal
processor inside the lens can ensure that the collected data is intervals on the left and right columns, as shown in Fig.3.
transmitted in real time after cloud computing. Through the
parallel connection of the lens, the system stability is improved
while ensuring high real-time performance.
The L-shaped calibration rod determines the origin of the
coordinate system of the entire system and the positive direc-
tions of the x, y, and z axes. And the T-shaped calibration rod
performs spatial calibration. The size of the calibration space
depends on the spatial range of the human body’s point of
movement. The Marker point is small, light and not easy to
damage, they are attached to the joints where the points need
to be collected.
C. 3D force measurement platform system
The 3D force measurement platform system mainly consists
of two force measurement platforms (as shown in Fig.2), NI
data acquisition card, and NK-Cortex software.
The 3D force measurement platform system mainly realizes
the mechanical force measurement and analysis functions of
the human body in the process of motion. The 3D force
sensor in the force measurement platform integrates the forces Fig. 3. Motion capture scene.
in the x, y, and z axes into multiple channel data during
measurement. And then the plantar force data acquisition by
the NI acquisition card, and finally the kinetic analysis of the B. Camera calibration
acquired data through the NK-Cortex software. Arrange 4 Marker points on the L-shaped calibration rod of
known size, and place them in the center of the motion capture
III. DATA COLLECTION AND PROCESSING OF HUMAN
field. Observe the position of the L-shaped calibration rod in
LOWER LIMB MOVEMENT
each lens through six lenses. The real physical size and pixel
A. Scene layout model parameters are estimated through the camera imaging
Before collecting experimental data, it needs to be arranged principle and computer vision geometric model.
on site. The cameras of the 3D motion capture system are In the experiment, it is necessary to adjust the three-
usually arranged according to the experimental requirements way pan/tilt of the corresponding lens, so that the L-shaped
[10]. Since the experiment done in this paper is to collect calibration rod is in the lower middle position of the lens. And
human lower limb movement data, the camera is placed in a adjust the angle of each lens, apron, etc., so that the number of
5 meters long and 4 wide In a space with a height of 3 meters Marker points in the lens is equal to the L-shaped calibration

6728

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY TIRUCHIRAPALLI. Downloaded on July 25,2024 at 08:55:11 UTC from IEEE Xplore. Restrictions apply.
 the number of 4 reflective marking points on the pole. After Trak or OpenSium software to analyze the changes in the
the L-shaped calibration rod is calibrated, remove it from the muscle force of the human body during exercise, so as to
capture area and take the T-shaped rod into the capture area do more in-depth research on it.
for calibration. The adjustment is mainly based on the size
E. Gait trajectory data processing
of the field and the size of the capture space. The operator
holds the T-shaped calibration rod and swings it repeatedly to
make its motion trajectory covers the entire capture area. After
calculation by the software, if the calibration passes, it means
that the motion capture in the scene is more accurate and data
can be colloected. If the calibration fails, it may be that there
are noises in the calibration process or the area covered by the
swing of the T-calibrator is not enough.

C. Marker points pasted

The 3D optical motion capture system used in this study
mainly captures the movement trajectory of the reflective
Marker points in space, so some reflective Marker points
need to be attached to the captured object when capturing.
In the process of recapture, the object captured is the Marker
point, not the experimenter, so it is more important for the
experimenter to wear the clothes. If the experimenter wears
loose clothes, the clothes shake severely during the capture
process and may cover up Marker point, which will affect the
capture effect. Therefore, this paper selects a healthy women Fig. 4. Imitating robot model.
with height of 165cm and weight of 56kg as the experimental
subject for exercise data collection. A total of 15 Marker Whether the collected human gait is good will directly affect
points are attached to both sides of the lower limbs of the the stability, coordination and driving torque of the lower
experimenter. During the experiment, the experimenter wore limb rehabilitation robot. The ideal gait is the continuous and
black leggings and black sports shoes, put hands on chest and smooth motion trajectory of each joint, such as position, angle,
walked naturally. angular velocity and angular acceleration. The 3D motion gait
capture system mainly captures the spatial position coordinate
D. Data collection of human lower limb movement information of the hip joint, knee joint and ankle joint. During
In this paper, the NK-Cortex software stores the collected the measurement process, due to occlusion or other external
human gait data in real time, and can not only perform factors, some marker point coordinate data is missing and
data processing and gait analysis on the collected human the curve is disconnected. The cortex capture software Make
movement data, but also establish a human skeleton model. a patch. In the acquisition process, due to the influence of
The NK-Cortex software can adjust the sampling frequency external factors such as the jitter of marker points and envi-
according to the research needs and the movement state of ronmental capture during human movement, it is necessary to
the Marker points can be seen in real time through it. When filter and fit the spliced data. Filtering methods mainly include
the human body moves in the space of the motion capture least squares filtering, wavelet filtering, Kalman filtering and
system, as long as the Marker points at the joints can be so on. In this paper, least square filtering and wavelet filtering
captured by two or more cameras at the same time, the spatial are used to process the data of the hip joint, knee joint and
position information of the Marker points at the moment can angle joint of the human left leg. Establish a lower limb
be recorded. exoskeleton robot model for the human body, as shown in
The quality of data collection plays an important role Fig.4.
in the research. Therefore, after the data is collected, it is The kinematics of the robot is used to study the geometric
necessary to check whether the position, velocity, acceleration relationship between the motion angle of each joint of the
and other information of each Marker point are missing. lower limb exoskeleton robot and its position in space. Simi-
Observe whether the curve is broken through the NK-Cortex larly, the relationship between the spatial position coordinates
software. If there is an obvious break in the middle of the of each joint of the human body’s lower limb and the angle of
curve, it means that the frame position data is lost. It must be each joint is found through the collected data. This relationship
repair the Marker points one by one. On the contrary, if the plays a vital role in the control of the gait trajectory data and
lost data fluctuates greatly or the number of consecutive lost the joint angle of the lower limb exoskeleton robot in the later
frames is too large, data collection needs to be repeated. A stage. Kinematics is divided into positive kinematics analysis
rigid body structure can be established through the repaired method and inverse kinematics analysis method. Based on the
data, and the collected data can also be imported into Outhro research needs, this paper adopts inverse kinematics analysis,

6729

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY TIRUCHIRAPALLI. Downloaded on July 25,2024 at 08:55:11 UTC from IEEE Xplore. Restrictions apply.
 that is, using the position data of the end of human lower limb cubic spline interpolation method for processing, and finally
hip jointa(ax , ay ), knee jointb(bx , by ), and ankle jointc(cx , cy ), processed the data with a frame rate of 128 frames per second.
the hip joint angle q1 and the knee joint angle q2 are obtained
by the tangent theorem,

q1 = arctan bbxy −a
−ay
x
(1)

α = arctan ccxy −b
−by
x
(2)

q2 = q1 − α = q1 − arctan ccxy −b
−by
x
(3)
From the results of the above two formulas and analyzing
the normal gait of the human
 body during
 walking,
 it can be
obtained q1 ∈ 0 π2 , q1 ∈ 0 π2 .Therefore, the
inverse kinematics equations of human body hip and knee
joints can be expressed as,

q1 = arctan bbxy −a
−ay
x

cx −bx (4)
q2 = q1 − arctan cy −by
Its motion trajectory is shown in Fig.5. Fig. 6. Human lower limb torque model.

The 3D force measurement platform system can measure

the 3D force data of the end position of the experimenters
lower limbs. Based on this article, the research is based on
two-dimensional plane. Therefore, the force in the x and y
directions is selected for research, as shown in Fig.6 , and
calculate the moment component of the end position, as shown
in Fig.7.
The component of the end position in the x direction is,

Mx = fx (cy − ay ) (5)
The component of the end position in the y direction is,

My = fy (cx − ax ) (6)
Through the Jacobian matrix, the torque at the end position
can be converted into the torque at the hip and knee joint
positions, as shown in Fig.8,
   
τ1  Mx
=J (7)
Fig. 5. The angle of hip joint and knee joint.
τ2 My
Where,
F. Plantar force data processing  
l1 cos θ1 + l2 cos (θ1 − θ2 ) −l2 cos (θ1 − θ2 )
The rhythm of the human body in the process of walking J=
−l1 sin θ1 − l2 sin (θ1 − θ2 ) sin (θ1 − θ2 )
is not fixed, so when the movement of the torque data is
(8)
too fast or too slow, the problem of under-sampling or over-
Therefore, the experimentalists hip joint torque is,
sampling will occur [11-12]. Over-sampling will lead to the
amount of information collected. Too much, which increases ∗
the burden of data analysis, and the utilization rate is not high; τ1 = (l1 cos θ1 + l2 cos (θ1 − θ2 )) fx (cy − ay )
∗ (9)
under-sampling will cause some important information to be + (−l1 sin θ1 − l2 sin (θ1 − θ2 )) fy (cx − ax )
lost, so this article oversampling the collected torque data. The experimentalists knee joint torque is,
Combined with the method of machine learning, the collected
100 groups of torque data are averaged. Aiming at the problem τ2 = − l2 cos (θ1 − θ2 ) ∗ fx (cy − ay )
(10)
of over-sampling of the processed data, this paper adopts the + sin (θ1 − θ2 ) fy (cx − ax )

6730

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY TIRUCHIRAPALLI. Downloaded on July 25,2024 at 08:55:11 UTC from IEEE Xplore. Restrictions apply.
 In the above formula l1 = 0.4m and l2 = 0.38m. B. Kinematics of human walking
The kinematics technology of the human body requires an
in-depth understanding of the basic walking of the human body
and the quantitative model of human walking[15]. One of the
main advantages of the established human walking quantitative
models is that they provide high-level parameters such as
speed, acceleration, and step length of each joint, which
can guide the generation of asynchronous states. Another
advantage is that the cost of this calculation is relatively low.
The goal of this method is to directly process the experimental
data obtained by the capture action. In forward kinematics, the
state vector of the articulated human lower limb movement
changes with time, and the interpolation method is often used
in the smoothing of the movement. In inverse kinematics,
Fig. 7. End force component of c. its algorithm can be used to solve some constraints, such
as foot penetration of the ground. Most of the kinematics
methods used to generate artificial synthetic motion rely on the
knowledge of biomechanics and combine forward kinematics
and inverse kinematics to calculate human motion.
C. The biomechanics of human walking
Human biomechanics focuses on the study of bone strength,
muscle strength, joint range and speed of various parts of the
human body. The purpose of human biomechanics research is
to make people play a better role in human-computer interac-
tion, try to avoid doing wasteful work, so that people wearing
lower extremity exoskeleton robots can receive effective re-
habilitation treatment, improve the efficiency of rehabilitation
training, and ensure personal safety[ 16]. On the other hand, in
Fig. 8. The torque of hip and knee Joint. order for the operator to wear the lower extremity exoskeleton
robot safely and comfortably, the design of the control system
must consider factors such as the size of the human body
IV. A NALYSIS OF HUMAN GAIT LAW output, the trajectory of the limbs, the smoothness of the
movement, and the direction of the movement of each part of
A. Gait cycle
the human body. Human biomechanics must be used. Related
Human walking movement plays a vital role in daily life. theories.
The study of human gait parameters during walking is of great In the process of walking, in order to find the clockwise
significance for understanding the laws of human motion and moment required by each joint of the human body in the
the coordination of limbs during walking [12]. Human walking process of walking, it is necessary to perform inverse dynamic
can also be described as a cyclic pattern of limb movement that analysis on the human body’s walking. Therefore, this paper
advances the body position. Assuming that all walking cycles records gait trajectory data and dynamic data through 3D
are the same, studying the walking process can be simplified motion capture system and 3D force measurement platform.
to studying the walking cycle. Through the analysis of the collected dynamics data, it is
When the human body is walking, the time interval between found that the vertical ground reaction force will be zero at the
two consecutive occurrences of the same action in a repetitive moment when the heel is in contact with the ground, and will
motion event becomes a gait cycle. Generally speaking, the rise sharply within a few tenths of a second. At the moment
walking cycle of the lower limbs of the human body is the foot is flat, the body mass moves down and onto the leg. In
composed of the support phase and the swing phase [13]. order to decelerate this downward movement and support the
Among them, the supporting phase is divided into single right body weight at the same time, a vertical force greater than the
supporting phase, single left supporting phase and double body weight needs to be applied to the foot. Fig.7 shows the
supporting phase. The swing phase is divided into a walking typical vertical ground reaction force profile of a single human
acceleration phase, a swing phase and a walking deceleration walking step. In the middle of standing, the movement of
phase. In a gait cycle of normal walking, the support phase the body’s center of mass is actually upward. This movement
accounts for about 60%, and the swing phase accounts for produces upward acceleration, which allows less than body
about 40%. With age, the proportion of support phase and weight to support the body. When the heel is off the ground,
swing phase in the gait cycle also changes[14]. the body mass is accelerated forwards and upwards, ready for

6731

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY TIRUCHIRAPALLI. Downloaded on July 25,2024 at 08:55:11 UTC from IEEE Xplore. Restrictions apply.
 the stand-up phase of the other leg, which means that more [13] X. Cheng, C. X. Wang, J. Wang, ea tl.,“ Research progress on treatment
than body weight is required to support the body. Finally, the of knee osteoarthritis from the perspective of lower limb biomechanics,”
Contemporary Medical Essays, no.7, pp. 124–125, 2014.
toe off the ground is the moment when contact with the ground [14] Y. P. Yi, “ The biomechanical characteristics of human gait and the
is lost and the force returns to zero. M-shaped diagram or clinical application of gait analysis,” Private Technology, no.4, pp. 83–
bimodal diagram is a typical display method of normal gait, 83, 2009.
[15] Z. H. Xu, J. W. Zhou, J. B. Ru “ Research and Implementation of Human
used to display force fluctuations. Motion Tracking System,” modern electronic technology, vol.29, no.29,
pp. 40–42, 2006.
V. C OUCLUSION [16] Z. X. Zhao, H. Dai, W. B. Zhao, “ Human motion posture simulation
based on inertial motion capture,” computer engineering, vol.38, no.5,
This article analyzes the 3D motion capture system and pp. 5–8, 2012.
the 3D force measurement platform system in detail; then
describes in detail the online collection and processing process
of human lower limb movement data, such as the layout of the
two sets of equipment scenes, equipment calibration, marker
point labeling, and collection process And the processing
process of gait trajectory data and gait force data; finally,
the collected human motion data is combined with its motion
mechanism to divide the gait cycle, kinematics analysis and
biomechanical analysis, based on the principle of biomedicine.
Therefore, researchers can better conduct research on rehabil-
itation training robots from the perspective of biomedicine.

ACKNOWLEDGMENT
The authors would like to thank the Young Backbone Teach-
er Training Program of Henan Provinces Higher Education
(2017GGJS117, 2018GGJS281), and National Natural Science
Foundation (U1813201) for their support of this work.

R EFERENCES
[1] C. P. Wang, B. X. Han, J. W. Luo, et al., “ Current status of exercise
rehabilitation therapy in rehabilitation of patients with stroke,” Advances
in clinical medicine, vol. 009, no.5, pp. 390–394, 2019.
[2] F. J. Badesa, R. Morales, N. Garciaaracil, et al., “ Dynamic Adaptive
System for Robot-Assisted Motion Rehabilitation,” IEEE Systems Jour-
nal, vol. 10, no.3, pp. 984–991, 2016.
[3] J. X. Zhang, Y. H. Cai, Q. Liu, et al., “ Research on the Influence
of Wearable Lower Limb Exoskeleton on Human Gait Characteristics,”
Journal of Biomedical Engineering, no.5, pp. 785–794, 2019.
[4] A. T. Abd, R. E. Singh, K. Iqbal and G. White, “ Muscle Synergies
are Robust across Participants in Upper Limb Rotational Motion,”
International Conference on Electrical and Electronics Engineering,
Antalya, Turkey, pp. 310–314, 2020.
[5] Y. Y. He, P. X. Li, and W. X. Yan, et al., “ Research on walking model
of lower limb assisted walking exoskeleton robot,” Control Engineering,
no.10, pp. 1803–1809, 2019.
[6] W. H. Xia, S. S. Zhang, “ Design of Motion Capture System Based on
Inertial Sensor,” Computer measurement and control, vol. 27, no.9, pp.
283–287, 2019.
[7] M. Silaghi, R. Pl?nkers, R. Boulic, et al., “ Local and global skeleton
fitting techniques for optical motion capture,” International Workshop
on Capture Techniques for Virtual Environments, pp. 26–40, 1998.
[8] H. Ren, D. X. Liu, N. Li, et al., “ On-line dynamic gait generation
model for wearable robot with users motion intention,” International
Conference on Information and Automation, pp. 347–352, 2018.
[9] T. B. Moeslund, A. Kruger, and K. Volker, “ A survey of advances in
vision-based human motion capture and analysis,” Computer Vision and
Image Understanding, vol.104, no.2, pp. 90–126, 2006.
[10] J. J. Zhang, F. M. Chen, “ Camera calibration in optical motion capture,”
computer application, no.S1, pp. 178–179+181, 2004.
[11] X. J. Wang, G. D. Guo, “ Comparative study of unbalanced data
sampling methods,” Microcomputer information, no.12, pp. 163–165,
2011.
[12] T. Yang, “ X. J. Wang, G. D. Guo, “ Comparative study of unbalanced
data sampling methods,” Nanjing University of Science and Technology,
pp. 6–30, 2014.

6732

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY TIRUCHIRAPALLI. Downloaded on July 25,2024 at 08:55:11 UTC from IEEE Xplore. Restrictions apply.