Visualization of Human Motion via Virtual Reality Interface and

Interaction based on It

Akihiro Sakurai

, Yosuke Ikegami

, Milutin Nikoli´c

, Yoshihiko Nakamura

and Ko Yamamoto

Graduate School of Information Science and Technology, University of Tokyo, Tokyo, Japan

Faculty of Technical Science, University of Novi Sad, Vojvodina, Serbia

Human-Motion Data Science Laboratory, Research into Artifact Center for Engineering, University of Tokyo, Tokyo, Japan

Keywords:

Virtual Reality, Motion Capture, Musculoskeletal Model, Sport Analysis.

Abstract: This paper presents the concept of ecological sports training and the initial developments based on biomechan-

ics analysis, VR technologies, and visualization of interaction. Computation algorithms in robotics have been

applied to biomechanics analyses including muscle force estimation. These technologies are widely used for

the self-biomechanics training, in which an athlete accesses and evaluates the analyzed results on his/her own

motions and will necessarily move toward the ecological training that considers interactions with the counter-

part and the environment in a sport game. We develop a VR visualization system of musculoskeletal analysis

that provides a realistic experience of the interactions to an athlete. We also report an initial evaluation of the

interactions with the virtual counterpart in the virtual environments through the VR system.

1 INTRODUCTION

Biomechanics is the basis of sports training. Tech-

nologies have been developed and widely used to fa-

cilitate the self-biomechanics training, which implies

a training where an athlete by oneself accesses scien-

tiﬁc evaluation and objective advice (Yoo et al., 2020)

(Maroto-Izquierdo et al., 2019). The analysis of mus-

culoskeletal activity exists as one of the methods of

motion analysis (Hill, 1938) (Stroeve, 1999) (Naka-

mura et al., 2005). And an application is being de-

veloped to visualize these results so that they can be

easily understood by non-specialists, subjects them-

selves (Murai et al., 2010). However, since many

sports are played not only by oneself but also by com-

peting against other competitors, in teams, or by in-

teracting with the surrounding environment, analysis

of only one subject is insufﬁcient. Therefore, it is

necessary to develop ecological training that includes

others surrounding the subject. Ecological training is

a type of training that involves not only oneself, but

also others and the environment proposed by Davids

and Handford (Davids et al., 1994) (Handford et al.,

1997). The possibility of stimulation from environ-

ments other than oneself is called affordance and is

proposed by Gibson (Gibson, 2014). By recreating an

environment with an affordance like that of a game,

the player can train with an understanding of not only

Figure 1: Visualization of musculoskeletal analysis on a

taekwondo kick motion measured by IMU-based motion

capture suite.

oneself but also the surroundings. In this study, we

deﬁne it as training that is conducted by presenting

others and environments other than oneself.

When ecological training is conducted, it is inad-

equate to reproduce the actual game or training en-

vironment by using conventional monitors for pre-

sentation. But the use of a VR headset during the

presentation increases the realism of the environment

and helps to build a training method that replicates

the environment surrounding the player. Reproduc-

ing the surrounding environment allows the subject

130

Sakurai, A., Ikegami, Y., Nikoli

c, M., Nakamura, Y. and Yamamoto, K.

Visualization of Human Motion via Virtual Reality Interface and Interaction based on It.

DOI: 10.5220/0010688800003059

In Proceedings of the 9th International Conference on Sport Sciences Research and Technology Support (icSPORTS 2021), pages 130-137

ISBN: 978-989-758-539-5; ISSN: 2184-3201

to play under the inﬂuence of them and us (including

the subject) to analyze their movements. Some stud-

ies of VR training and analysis has been conducted by

recreating environment. In sports, analysis of the dif-

ference in skills between advanced and novice play-

ers has been conducted, as well as controlled exper-

iments on the level of improvement when using VR

and when not using VR (Hosp et al., 2020) (Petri

et al., 2019) (Kimura et al., 2018). In other ﬁelds, VR

simulators for heavy machinery such as construction

site cranes and forklifts are being developed (Tech-

nologies, 2021), and research is being conducted on

drilling drill proﬁciency training (Zhang, 2017).

In order to support the ecological training method,

this study aims to develop a training method in which

the player is presented with a space that reproduces

the competition opponent and the surrounding envi-

ronment using VR, and the player can compete under

the inﬂuence of each other and analyze the movement.

In this case, the results of musculoskeletal analysis,

which is the result of biomechanical analysis of the

athlete, can be presented for use in ecological train-

ing. In this report, we developed a method to visu-

alize the results of biomechanical analysis presented

to players. In addition, we veriﬁed the reproducibility

of the interaction with the opponent and surrounding

environment in the VR space compared to the real en-

vironment by bidirectional walking.

2 MUSCULOSKELETAL MODEL

VISUALIZATION

2.1 Visualization Flow

In this study, we use a wire approximation model of

muscle (Nakamura et al., 2005). This model uses in-

verse kinematics to calculate the position and posture

of the skeleton from the 3D positions of markers that

represent the positions of body parts. The muscle

wires are basically attached to the skeleton, and from

the obtained skeletal motion and mechanical parame-

ters, the wire tensions are optimized using inverse dy-

namics, assuming that the sum of the squares of the

wire tensions is minimized.

The musculoskeletal rendering method using this

wire model (Murai et al., 2009) was based on

OpenGL, which doesn’t allow for fast and detailed

real-time rendering and display in VR headset. To

solve this problem, we implemented a fast-rendering

method using Unity, which is a development environ-

ment that supports a variety of platforms. In addition,

in the real-time rendering of muscles, polygon rupture

CPU

Musculoskeletal analysis

Forward Kinematics

Muscle raw shaping

Interpolating

Coloring

Joint angles

Muscle tensions

Image output

Muscle activity

GPU

Geometry shader

Wire vertices

Figure 2: Musculoskeletal visualize ﬂow. Polygon position

calculation is divided to CPU part and GPU part.

occurs when rendering diffraction shapes, causing a

sense of discomfort, but we have solved this prob-

lem by rendering detailed muscle shapes. And by us-

ing geometry shader which increases the number of

polygons in GPU simultaneously, we distribute cost

between the CPU and GPU and improve processing

speed. The ﬂow of the visualization process is shown

in Figure. 2. Joint angles and muscle tensions are

obtained from the results of musculoskeletal analysis.

From the joint angles, the wire vertices of the mus-

cle can be obtained by solving forward kinematics on

the CPU, which are input to the GPU to calculate the

muscle shape using a geometry shader. Muscle ten-

sions are used as input to the GPU for coloring them.

The ﬁnal screen output is obtained through the above

processing.

2.2 Creation of Muscle Shape Polygons

We create a muscle-like appearance while avoiding

tearing of the diffraction shape by using the wire mus-

cle model (Nakamura et al., 2005). Let the wire ver-

tices of the muscle be p

, p

,·· · , p

n−1

in order. p

the origin point and p

n−1

is the arrest point. The other

vertices are deﬁned as passing points. Obtain a circle

orthogonal to the line connecting p

and p

i+1

. C

represented by two vectors u

and v

that are orthogo-

nal to each other.

t = p

i+1

− p

(t = (t

)) (1)

= (0,t

,−t

) (2)

= (t

+ t

,−t

) (3)

(4)

The radius of this C

is changed to create the spin-

dle shape of the muscle. To reproduce the spindle

Visualization of Human Motion via Virtual Reality Interface and Interaction based on It

131

shape of the muscle, the radius is calculated using the

quadratic function.

f(x) =

− a

)

(5)

f(0) = pa

/4 is the maximum radius of the mus-

cle, which is proportional to the maximum force of

the muscle. The spindle shape is created as shown

in Figure. 3. If one of the two points representing

the line l

is the origin or arrest point, a half spindle-

shaped polygon is created; otherwise, a cylindrical

polygon is created. At this time, there is a gap be-

tween the polygons near the transit point due to the

diffraction shape as shown in Figure. 4.

To ﬁll this gap, we create new triangular polygons

by connecting the polygons on the two orthogonal

circles on the passing point. Let the two orthogo-

nal circles on the passing point be C

and C

i+1

, and

the unit vectors representing each circle be u

and

i+1

. At this time, the degree of the polygon gap

changes depending on the bending angle of the wire

on the target passing point. Therefore, the interpola-

tion method is changed according to the bending an-

gle, and appropriate interpolation is performed.

Figure 3: Shuttle 3d polygon points. Each circle radius is

due to f(x) = p(x

− a

)

/4.

2.2.1 When the Bending Angles is 120

◦

or More

Let the radius of the orthogonal circle be r. At this

time, the polygon vertices on C

are obtained when

Cylindrical shape

Spindal shapes

Polygon gaps

Figure 4: Polygon gap by

the wire bending. Green

line shows wired muscle.

Red triangles and rect show

created muscle polygons.

Interpolatring

polygons

Spindal

shape

Figure 5: Polygon gap on

the passing point is interpo-

lated.

the circles are approximated by regular n-gon.

= p

+ r(u

cos2π

+ v

sin2π

) (0 ≤ j < n)

(6)

Select the vertices t

k,0

k,1

k,2

, which consist the

triangle polygon t

, from on each orthogonal circle so

that the each distance between the polygon vertices is

minimized.

k,0

= q

i, j

(7)

k,1

= q

i+1, j+1

(8)

k,2

(

i+1, j

(0 ≤ k < n)

i, j+1

(n ≤ k < 2n)

(9)

This selection of vertices interpolates the gap with

a total of 2n triangular polygons as shown in Figure.

t t

Figure 6: Polygon gap creation from orthogonal circles on

the passing point. Each triangle has 3 polygon vertices on

the circles.

2.2.2 When the Bending Angle Is Less than 120

◦

When the bending angle is 120

◦

or more, a triangular

polygon was constructed by connecting the polygon

vertices on the C

and C

i+1

to hide the crack, but if

the angle is close to an acute angle, the interpolation

is rapid and unnatural. So, we create an intermediate

circle C

and perform a gentle interpolation. Let l

and l

i+1

be the lines that make up the wire that inter-

sects at the passing point and let n

and n

i+1

be the di-

rectional vectors on the lines. The unit vectors u

that represent the orthogonal circle C

which divides

the wire segment are obtained as follows. Interpola-

tion is by ﬁlling two gaps betweenC

and C

, C

, and

i+1

= n

+ n

i+1

(10)

(11)

= n

× n

i+1

(12)

2.3 Fast Rendering by Geometry

Shader

In this way, natural muscle polygons with no tears can

be created as shown in Figure. 5. The process of cre-

icSPORTS 2021 - 9th International Conference on Sport Sciences Research and Technology Support

132

ating muscle polygons is performed using the geome-

try shader on GPU. The geometry shader is a process

that takes polygon vertices as input and increases the

number of vertices. Parallel processing on the GPU

allows for fast polygon increment processing.

3 SPORT MOTION

PRESENTATION ON VR

WORLD

3.1 Motion Measurement and

Musculoskeletal Analysis

We conducted measurement and presentation to re-

produce the interaction in VR world with an opponent

by taekwondo motion, which is one of the interper-

sonal sports. We measured taekwondo motion of one

player and one coach by IMU-suit motion capture and

applied musculoskeletal analysis. The obtained mus-

culoskeletal motion was presented on a head-mounted

display to taekwondo athletes, coaches, who have

knowledge of taekwondo for evaluation. The IMU

suit was used in the measurement experiment of taek-

wondo. The result of musculoskeletal analysis by ap-

plying inverse kinematics and inverse dynamics to the

measured motion is shown in Figure. 1.

3.2 Presentation on VR World

The measured motion is visualized in the VR world

using the musculoskeletal analysis result drawing

method described in Chapter 2 and presented to the

subjects wearing the VR headset. The head-mounted

display used for the presentation was the Oculus

Quest2 by Oculus. The subjective evaluation and

response of the subjects during the presentation are

shown below.

• The musculoskeletal presentation using a display

was not so good, but the VR presentation using

headset had a completely different sense of scale

and gave a sense of realism to the kicking.

• You could observe the motions up close, and al-

though you could see the differences in the kicks

of the players on the display, I could see the dif-

ferences even more.

• The subjects felt that measuring the opponent’s

kicks in competitions and presenting the mea-

sured motions in VR would be more useful than

research using displays to learn the opponent in

advance.

3.3 Discussion

From the impressions and evaluations, it is consid-

ered that the musculoskeletal motion presentation can

reproduce the interaction with the opponent as in the

real world. By using this interaction, we can repro-

duce the movements of the opponent, which is dif-

ﬁcult to achieve in the real world, and lead to more

efﬁcient training than before.

4 MEASUREMENT

INTERACTION ON REAL

WORLD

To reproduce human-human interaction in the VR

world, we conducted 3D integrated measurement ex-

periments of humans and the environment. The mo-

tion obtained from the musculoskeletal analysis of the

measured data is presented in the VR space.

4.1 Measurement Methods

4.1.1 Motion Capture

In this experiment, we use video motion capture

(VMocap) developed by Ohashi et al. (Ohashi et al.,

2018), which estimates the 3D positions of joints by

capturing images with multiple cameras and using the

probability of the existence of the subject’s joints ob-

tained by Openpose (Cao et al., 2017).

4.1.2 Environment Measurement

VisualSLAM (Moulon et al., 2012) is a method for

estimating the 3D positional relationship between the

camera that captured the image and the feature points

based on the correspondence between images of fea-

ture points on the image, using multiple images of the

object as input. The application of photogrammetry,

Meshroom (AliceVision, 2018), can be used to recon-

struct 3D models of the environment.

4.2 Conditions

The measurement space is shown in Figure. 7, where

there is an obstacle like a desk in the center of the

environment, and subjects passing near the obstacle

and interacting with the environment. The 3D model

of the measurement environment was created using

Meshroom in advance, and the measurement was per-

formed in a situation where subjects were passing

each other, and VMocap was used to obtain the mo-

tion of the subjects. We instructed two subjects to

Visualization of Human Motion via Virtual Reality Interface and Interaction based on It

133

Recording space

Desk

Kitchen

Entrance

Point A

Camera direction

Figure 7: Recording area, which has entrance and desk at

the center. 7 cameras covered area.

perform the following walking motions in a real envi-

ronment and asked them to pass each other.

• The subject who walks clockwise around the desk

from the entrance is designated as Subject A. Sub-

ject A is asked to pick up the pencil case on the

desk and proceed to the entrance again.

• The subject who passes subject A is called subject

B. Subject B was given no instructions other than

to pass each other.

4.3 Reconstruction Result

Figure. 8 is the result of subject’s key points detec-

tion by Openpose. Key points are well detected in

the environment. The musculoskeletal motion and en-

vironment estimation results of two subjects passing

motion is shown in Figure. 9.

Figure 8: Subject key points estimation result.

Figure 9: Musculoskeletal motions of the passing motion in

the environment.

5 INTERACTION

REPRODUCTION ON VR

WORLD

5.1 Presentation Method

The data measured in the previous chapter are repro-

duced in the VR world, and the presentation experi-

ment is conducted to reproduce the passing motion of

the subject in the real space and the VR world. For

the presentation, we used the Oculus Quest2.

5.2 Conditions

The subjects are asked to reproduce the following ac-

tions in the real world measured in the previous sec-

tion.

1. The subject lonely starts at the entrance, moves

clockwise around the desk in the center of the en-

vironment, retrieves the pencil case, and returns

to the entrance.

2. Passing motion with a person starting from the

Point A and proceeding to the washroom.

3. Passing motion with a person starting from the en-

trance and proceeding clockwise around the desk

Since Motion 1 is performed by only one person, the

VR presentation does not use the motion measured in

the previous section. In order to compare whether the

behavior changes depending on the presentation of

the person passing by, we used the presentation with

muscle rendering and the presentation with only the

skeleton without muscle rendering in Motion 2 and

3. The conditions used in this study are the following

motions measured in the previous section.

a. Motion 2 with muscle drawing

b. Motion 2 without muscle drawing

c. Motion 3 with muscle drawing

d. Motion 3 without muscle drawing

A total of ﬁve subjects participated in the study.

5.3 Results

5.3.1 Trajectory Comparison

Figure. 10 is the walking motion of a single person

moving clockwise around a desk. The red line shows

the trajectory in real world, blue lines show the tra-

jectory in VR world. Figure. 11 is a superimposed

drawing of the walking motion in the VR world and

the walking motion in the real world under conditions

a, b, c and d.

icSPORTS 2021 - 9th International Conference on Sport Sciences Research and Technology Support

134

Figure 10: Trajectories of walking around desk (Motion 1).

Red line shows trajectory measured in real world by video

motion capture. Blue lines show subjects trajectory.

Walk Motion a) Walk Motion b)

Walk Motion c) Walk Motion d)

Figure 11: Trajectories of crossing paths with Opponents

a), b), c), d).

In the clockwise walking motion in the environ-

ment (Motion 1), the trajectory in the real world pro-

gresses smoothly around a corner, while the trajectory

in the VR world tends to make a right turn. The trajec-

tory in the VR world remains in line with the desk for

most subjects, while the trajectory in the real world

leaves the desk immediately after the acquisition of

the pencil case. Comparing the trajectories in the VR

space in condition a and b, the trajectories in condi-

tion a showed less variation. In the action of walking

pass each other while retrieving the pencil case (Mo-

tion 2), the trajectories in the real world and in the VR

world are almost the same in the Motion 3 of passing

a person who acquires a pencil case, especially in the

middle of the trajectory where they pass each other.

The trajectories in the VR space for condition c and

d were compared, and as in the case of Motion 2, the

trajectories in condition c showed less variation.

5.3.2 Distance and Gaze Direction Vector

against Opponent

We focused the change on the distance between the

subject and the passing opponent as an interaction.

Figure. 12 is a time series graph of the distance be-

tween the subject’s head and the passing opponent

projected onto a horizontal plane. The vertical line in-

dicates the timing of the minimum distance between

the subject and the passer-by. View the distance graph

in the beginning. In all conditions, the timing of the

minimum distance to the opponent is faster in the case

of passing each other in the real world. In addition, in

conditions a and b, the minimum distance to the op-

ponent tends to be smaller in the case of passing each

other in the VR world.

Next, we focused on the gaze direction graph. Fig-

ure. 13 is a time series graph of the gaze vector angle

against the passing opponent. The closer the angle is

to 0

◦

, the more the subject is gazing at the opponent.

In all the ﬁgures, we can see that the subjects in the

VR world gaze at the opponent longer. In addition, in

conditions a and b, the subjects in the real world do

not gaze at the other person for a long time, such as

immediately removing their gaze after gazing at the

other person for about one second from the beginning

of passing, while the subjects in the VR world tend to

gaze at the other person until just before passing.

5.4 Discussion

Using the walking motion in the real world measured

in the previous section and the environment informa-

tion reconstructed by VisualSLAM, we conducted a

reproduction experiment in the VR world to compare

the walking motion in the real world and the walking

motion in the VR world. Awkward clockwise walk-

ing in VR world (Motion 1) is thought to be because

the subject did not fully grasp of the space in the VR

world and felt resistance to moving while walking. In

addition to grasping of the space, the actual environ-

ment in which the VR reproduction experiment was

conducted was surrounded by walls, and the subject

is controlled not to collide with such real obstacles.

This movement restrictions are also shown in the time

series graphs of the distance and direction of gaze of

the passers-by, and the time at which large changes

occur in the graphs is often later in the VR world than

in the real world.

When we compared the time when the opponent in

VR world was visualized with muscles included and

the time when only the skeleton was visualized, the

variation of the path was less when muscles were in-

cluded. The reason for this is that including muscles

Visualization of Human Motion via Virtual Reality Interface and Interaction based on It

135

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

Distance between heads[m]

dis.1

dis.2

dis.3

dis.4

dis.5

dis.ref

Walk Motion a

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

Distance between heads[m]

dis.1

dis.2

dis.3

dis.4

dis.5

dis.ref

Walk Motion b

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

Distance between heads[m]

dis.1

dis.2

dis.3

dis.4

dis.5

dis.ref

Walk Motion c

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

Distance between heads[m]

dis.1

dis.2

dis.3

dis.4

dis.5

dis.ref

Walk Motion d

Figure 12: Solid lines show time series change of distance

between subject and opponent. Dotted vertical lines show

the timing of minimum distance to the opponent.

increased the existence of the opponent and narrowed

the range of the subject’s choice of path. In addition,

the subjects thought that they were passing between

a desk and the opponent in the environment, and the

environmental information was given close to the re-

ality. Although the presence or absence of the mus-

cle visualizing inﬂuenced the choice of the path, the

minimum distance to the passing partner was not dif-

ferent. It did not inﬂuence the timing of the actual

crossing, when the opponent is reﬂected at the edge

of the ﬁeld of vision, making it more difﬁcult to see

than in the center of the ﬁeld of vision. When we

compared how long the subjects gazed at the oppo-

nent, the subjects in the VR world tended to gaze at

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

100

150

Gaze direction vector angle[°]

gaze dir.1

gaze dir.2

gaze dir.3

gaze dir.4

gaze dir.5

gaze dir.ref

Walk Motion a

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

100

150

Gaze direction vector angle[°]

gaze dir.1

gaze dir.2

gaze dir.3

gaze dir.4

gaze dir.5

gaze dir.ref

Walk Motion b

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

100

150

Gaze direction vector angle[°]

gaze dir.1

gaze dir.2

gaze dir.3

gaze dir.4

gaze dir.5

gaze dir.ref

Walk Motion c

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Time[s]

100

150

Gaze direction vector angle[°]

gaze dir.1

gaze dir.2

gaze dir.3

gaze dir.4

gaze dir.5

gaze dir.ref

Walk Motion d

Figure 13: Solid lines show time series change of angle be-

tween the line of sight and the line connecting subject and

opponent. Dotted vertical lines show the timing of mini-

mum distance to the opponent.

him longer. It is thought that it was necessary to ob-

serve him for a long time when passing each other

in the VR world. First, unlike in the real world, the

motion of the passer-by is based on a musculoskeletal

model, so it is easy to attract the eyes of the subjects.

In addition, visualizing the skeletal makes it difﬁcult

to see the eyes of the opponent and to know which di-

rection he is facing and moving to. So, the estimated

results of muscle tension did not affect the prediction

of the movement of the opponent and the difﬁculty in

predicting the direction of gaze may have affected the

subject’s walking.

icSPORTS 2021 - 9th International Conference on Sport Sciences Research and Technology Support

136

On the other hand, the presentation of environ-

mental information by VisualSLAM does not give the

subject a sense of discomfort, but is given as infor-

mation, and can play a role close to reality in the in-

teraction between the subject, the opponent, and the

environment. For sport training, it makes a sense to

use the musculoskeletal model. But for reproducing

real environmentin VR world, it is assumed that using

more human-like models is more important. It is as-

sumed that it is more important to use models that do

not give the subject a sense of discomfort while keep-

ing the joint positions close to the actual one, rather

than using a musculoskeletal model for the presenta-

tion of the passing partner, in order to reproduce the

passing gait in the VR world, not in sport training.

6 CONCLUSION

The results of musculoskeletal analysis were rendered

in VR space, which is useful for ecological training in

interpersonal sports. The environmental information

was well reproduced, but the general human model

was more effective in the walking motion experiment.

In the future, it is necessary to reproduce the subject’s

own information in order to use it for ecological train-

ing.

REFERENCES

AliceVision (2018). Meshroom: A 3d reconstruction soft-

ware. https://github.com/alicevision/meshroom.

Cao, Z., Simon, T., Wei, S.-E., and Sheikh, Y. (2017). Real-

time multi-person 2d pose estimation using part afﬁn-

ity ﬁelds. In Proceedings of the IEEE conference on

computer vision and pattern recognition, pages 7291–

7299.

Davids, K., Handford, C., and Williams, M. (1994). The

natural physical alternative to cognitive theories of

motor behaviour: An invitation for interdisciplinary

research in sports science? Journal of sports Sciences,

12(6):495–528.

Gibson, J. J. (2014). The ecological approach to visual per-

ception: classic edition. Psychology Press.

Handford, C., Davids, K., Bennett, S., and Button, C.

(1997). Skill acquisition in sport: Some applications

of an evolving practice ecology. Journal of sports sci-

ences, 15(6):621–640.

Hill, A. V. (1938). The heat of shortening and the

dynamic constants of muscle. Proceedings of the

Royal Society of London. Series B-Biological Sci-

ences, 126(843):136–195.

Hosp, B., Schultz, F., H¨oner, O., and Kasneci, E. (2020).

Eye movement feature classiﬁcation for soccer goal-

keeper expertise identiﬁcation in virtual reality. arXiv

preprint arXiv:2009.11676.

Kimura, T., Nasu, D., and Kashino, M. (2018). Utilizing

virtual reality to understand athletic performance and

underlying sensorimotor processing. In Multidisci-

plinary Digital Publishing Institute Proceedings, vol-

ume 2, page 299.

Maroto-Izquierdo, S., Fernandez-Gonzalo, R., Magdi,

H. R., Manzano-Rodriguez, S., Gonz´alez-Gallego, J.,

and De Paz, J. a. (2019). Comparison of the mus-

culoskeletal effects of different iso-inertial resistance

training modalities: Flywheel vs. electric-motor. Eu-

ropean journal of sport science, 19(9):1184–1194.

Moulon, P., Monasse, P., and Marlet, R. (2012). Adaptive

structure from motion with a contrario model estima-

tion. In Proceedings of the Asian Computer Vision

Conference (ACCV 2012), pages 257–270. Springer

Berlin Heidelberg.

Murai, A., Kurosaki, K., Yamane, K., and Nakamura, Y.

(2009). Computationally fast estimation of muscle

tension for realtime bio-feedback. In 2009 Annual

International Conference of the IEEE Engineering

in Medicine and Biology Society, pages 6546–6549.

IEEE.

Murai, A., Kurosaki, K., Yamane, K., and Nakamura, Y.

(2010). Musculoskeletal-see-through mirror: Compu-

tational modeling and algorithm for whole-body mus-

cle activity visualization in real time. Progress in bio-

physics and molecular biology, 103(2-3):310–317.

Nakamura, Y., Yamane, K., Fujita, Y., and Suzuki, I.

(2005). Somatosensory computation for man-machine

interface from motion-capture data and musculoskele-

tal human model. IEEE Transactions on Robotics,

21(1):58–66.

Ohashi, T., Ikegami, Y., Yamamoto, K., Takano, W., and

Nakamura, Y. (2018). Video motion capture from the

part conﬁdence maps of multi-camera images by spa-

tiotemporal ﬁltering using the human skeletal model.

In 2018 IEEE/RSJ International Conference on Intel-

ligent Robots and Systems (IROS), pages 4226–4231.

IEEE.

Petri, K., Emmermacher, P., Danneberg, M., Masik, S.,

Eckardt, F., Weichelt, S., Bandow, N., and Witte, K.

(2019). Training using virtual reality improves re-

sponse behavior in karate kumite. Sports Engineering,

22(1):2.

Stroeve, S. (1999). Impedance characteristics of a neuro-

musculoskeletal model of the human arm i. posture

control. Biological cybernetics, 81(5):475–494.

Technologies, I. (2021). Mining training simulators.

https://www.immersivetechnologies.com/products/

Simulators.htm.

Yoo, K., Wu, X., Zhuang, W., Xia, Z., and Liu, Y. (2020).

The effects of audible feedback as a coaching strategy

on golf skill learning for novice players. International

Journal of Performance Analysis in Sport, 20(4):596–

609.

Zhang, H. (2017). Head-mounted display-based intuitive

virtual reality training system for the mining industry.

International Journal of Mining Science and Technol-

ogy, 27(4):717–722.

Visualization of Human Motion via Virtual Reality Interface and Interaction based on It

137