Analysis of Three-Dimensional Human Movements and Its Research

in Sports Dance Teaching

Ruili Zheng

1,*

and Jingqi Zhang

Physical Education Research Department, Xinjiang University, Urumqi, Xinjiang, China

Office of the president of the Party Committee, Xinjiang Institute of Engineering, Urumqi, Xinjiang, China

Keywords: VR Technology, Three-Dimensional Analysis

, Human Movement, Movement Analysis, Dance Teaching

Abstract: Since human beings are the most important element in the natural and social environment, various human

activities are accompanied by a wealth of important information about the interaction between humans and

nature, and between humans and society. The automatic learning and recognition of human action behavior

is of great significance in the fields of advanced human-computer interaction, automatic monitoring of human

behavior, and intelligent analysis of sports, and has a wide range of application prospects. This article mainly

studies the three-dimensional human movement analysis and its application in sports dance teaching. This

article summarizes three methods for obtaining 3D human action skeleton data, and uses depth map sequence

as the method for obtaining action skeleton data in this article. Then, the 3D human skeleton model and action

pose feature description method in this paper are proposed. This article describes the key frame extraction

method of 3D human motion data, summarizes the characteristics of 3D human motion data, and analyzes the

current main key frame extraction methods. This paper studies the 3D human action recognition methods and

analyzes the advantages and disadvantages of various recognition methods. This article studies the current

state of sports dance teaching, and based on this, proposes a teaching method that combines sports dance with

three-dimensional human movement analysis. Through experiments, it is found that when analyzing the

movements in the dance process, the estimation errors of the human head, pelvis, upper and lower arms and

upper and lower legs in the first 300 frames are all under 70mm, which can accurately analyze the current

dance movements. The standardization can help students learn dance better.

1 INTRODUCTION

With the vigorous development of our country's

economy and the continuous acceleration of the pace

of people's lives. Faced with the ever-increasing

pressure of life, people devote more time to work and

study, while the time allocated to physical exercise

and dance art is getting less and less, which invisibly

leads to a general decline in the physical fitness of the

public (Purvis and Denise, 2016). With the rapid

development of the electronic information industry,

more and more electronic products are applied to all

areas of life (Wei, 2017). In this article, a sports dance

teaching method using a three-dimensional human

movement analysis instrument is proposed to help

students learn dance.

Many scholars have conducted research and

analysis on three-dimensional human movement

analysis and its application in sports dance teaching.

For example, Ayame Yamazaki, Takuo Ikeda and

Takeshi Tsutsumi applied video fundus photography

technology to three-dimensional recording and

analysis of eye movements, which required

quantification of the main sequence (Borgogno,

2017). They used infrared images to obtain the

characteristics of the torsion fundus and analyzed its

main sequence (Hooshang and Eric, et al. 2016). Guo

Yusun, Chen Wenjuan and others proposed a new

dance self-learning framework based on Laban's

motion analysis principles (Yamazaki and Ikeda, et

al. 2019), so that students can automatically analyze

dance movements and correct dance skills without an

expert (Sun and Chen , et al. 2017). They proposed a

"shape-effort" feature description model to reflect the

subtleties of dance movement.

This article introduces the related technologies of

3D human motion analysis in detail, including 3D

human motion skeleton data acquisition method,

motion key posture frame extraction method and

human motion recognition (Wang and Huang, et al.

Zheng, R. and Zhang, J.

Analysis of Three-Dimensional Human Movements and Its Research in Sports Dance Teaching.

DOI: 10.5220/0013538700004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 1, pages 227-232

ISBN: 978-989-758-763-4

227

2016). This article introduces the current situation of

sports dance teaching, and puts forward a sports

dance teaching method based on three-dimensional

human movement analysis technology (Jiang and

Zhang , et al. 2017). This article analyzes the error of

each limb joint when the three-dimensional human

body movement analysis instrument analyzes sports

dance movements through experiments (Migliorati

and Cevidanes , et al. 2021). This article analyzes the

effect of three-dimensional human movement

analysis on sports dance teaching through

experiments.

2 RESEARCH ON THE ANALYSIS

OF 3D HUMAN MOVEMENT

CHARACTERISTICS BASED

ON VR TECHNOLOGY AND

ITS APPLICATION IN THE

DESIGN OF SPORTS DANCE

TEACHING SYSTEM

2.1 3D Human Motion Analysis

Related Technology

2.1.1 Three-dimensional Human Body

Motion Skeleton Data Acquisition

Method

In computer vision, there are three main methods to

obtain 3D human action skeleton sequence: based on

multi-view 2D video image sequence reconstruction,

based on 3D motion capture system acquisition and

based on depth map sequence mapping (Shogo and

Yasuhiro, et al. 2018). Due to differences in human

body shape, lack of depth information in 2D images,

and partial self-occlusion, it is difficult to accurately

estimate 3D human bones (Fan and Zheng, et al.

2018). At present, the 3D skeletal joint data of the

human body is mainly obtained by two methods

based on the 3D motion capture system and the depth

map sequence. The 3D skeletal joint data obtained

based on the motion capture system has higher

accuracy and fewer noise points, but the motion

capture equipment is expensive, cumbersome to use

and generally not applicable (Peng C and Pan B Z, et

al. 2020). Based on the depth information collection

and mapping methods of the depth sensor, the

prediction of 3D bone joints usually has errors, and

the depth map will also contain noise, but the depth

sensor is small in size and has universal applicability.

2.1.2 Action Key Pose Frame

Extraction

Method

The key posture framework of the three-dimensional

human body action refers to the posture that can best

reflect the action changes in the action and

represented by the 3D bone joint coordinate data. The

current 3D motion data key frame extraction methods

are mainly divided into two types: uniform sampling

extraction and adaptive sampling extraction. Uniform

sampling extraction refers to re-sampling the motion

sequence at equal time intervals. Due to the problems

of undersampling and oversampling (leading to

missing and redundant key frames), this method has

not been widely used. The method of adaptive

sampling to extract key frames usually uses the

original motion data to be converted into motion

feature description, and automatically extracts the

posture of the key frame by analyzing the motion

posture feature of the action posture sequence, which

solves the uniformity problem well. At present, the

adaptive sampling and extraction of key frames are

mainly divided into three categories: frame

subtraction, curve simplification and clustering.

2.1.3 Human

Action Recognition

Human action recognition research belongs to the

category of pattern recognition. After describing the

mathematical model of action posture features, it

mainly includes two basic tasks: standard action

classifier design and action classification recognition.

According to the characteristics of the algorithm,

action recognition algorithms are mainly divided into

three categories: methods based on template

matching, methods based on state space, and methods

based on syntax analysis. The template-based method

is easy to implement, does not require a large number

of training action samples, has a small amount of

calculation, and has a higher recognition rate when

the quality and parameters of the reference template

are both optimized. However, this method is sensitive

to the length of the action gesture sequence and noise

points, and its robustness is not good enough. It is

usually suitable for the classification and recognition

of simple actions. The method based on state space

can effectively overcome the problem that template

matching is sensitive to noise. The algorithm has high

robustness and can recognize simple and continuous

actions. It is the current mainstream action

recognition method and is widely used. However, this

method also has disadvantages. In order to obtain an

ideal classifier model, a large number of action

samples are required for training. For a classifier

INCOFT 2025 - International Conference on Futuristic Technology

228

model with more parameters, the amount of

calculation is relatively large. The method based on

grammatical analysis is helpful to understand the

complex action structure and effectively use the prior

knowledge, but the current research is still in its

infancy, usually combined with the first two methods.

2.2 B.VR Technology

Virtual reality technology is developed by integrating

multiple technologies, including real-time 3D

computer graphics technology, wide-angle stereo

display technology and head tracking technology. In

the VR system, the image that the user sees through

the VR device depends on the position and direction

of the eyes, but the human visual system and the

motion system are separated, which is determined by

the structure of the human brain. Using head tracking

technology can link the vision system with the motion

system, so that the virtual objects in the VR system

look more realistic.

2.3 Sports Dance Teaching

In the past, when people wanted to learn sports dance,

they usually had two choices: go to a sports dance

club to learn from a teacher or learn by themselves by

watching animations and videos.

2.3.1 Sports Dance Club

At present, the most commonly used method for

dance teaching is the demonstration-practice method.

Demonstration-exercise is a simple and reasonable

teaching method that can help people learn movement

and mental skills. In this method, the teacher must

first demonstrate the dance for the students, and then

the students imitate the teacher's movements under

the teacher's on-site supervision. Afterwards, the

teacher will give feedback to inform the students of

their performance, thereby helping the students to

further improve their dance moves. Traditionally,

dance demonstrations and information feedback must

be completed by a teacher. This learning model has

proven to be effective, but the dependence on the

teacher makes this learning model lack of

convenience and flexibility.

2.3.2 Sports Dance Teaching System

With the development of technology, students can

now take sports dance courses without the teacher

present. Some sports dance teaching systems have

been developed and put into use one after another.

However, these systems can only provide simple

demonstration functions, and they cannot provide any

feedback information to help students improve their

sports dance moves.

2.3.3 Sports Dance Class Using Three-

Dimensional Human Movement

Analysis

According to the current situation of sports dance

teaching, the idea of combining three-dimensional

human movement analysis technology with sports

dance teaching system and applying it in the sports

dance classroom was put forward, and a comparative

experiment was done. Three-dimensional human

movement analysis can more accurately analyze the

current standardization of dance movements, which

can help students learn dance better. The application

of three-dimensional human movement analysis in

sports dance classrooms helps to increase students'

interest in sports dance, helps improve students'

learning efficiency, and enables students' subjectivity

to be brought into full play, thereby improving

teaching efficiency and quality.

3 ANALYSIS OF 3D HUMAN

MOVEMENT

CHARACTERISTICS BASED

ON VR TECHNOLOGY AND

ITS APPLICATION

EXPERIMENT IN THE DESIGN

OF SPORTS DANCE

TEACHING SYSTEM

3.1 Three-Dimensional Human

Movement Analysis-Error Analysis

3.1.1 Test subject

This article collects 10 dance talents as volunteers to

conduct experiments. The three-dimensional human

motion analysis instrument records the physical state

of the 10 volunteers when they dance the same dance,

analyzes the difference between the data recorded by

the three-dimensional human motion analysis

instrument and the actual data, and understands the

accuracy of the three-dimensional human motion

analysis instrument in analyzing dance movements.

Analysis of Three-Dimensional Human Movements and Its Research in Sports Dance Teaching

229

3.2 Evaluation Experiment of Dance

Class Using Three-Dimensional

Human Movement Analysis

3.2.1 Experimental Setup

A total of 46 students, including 29 girls and 17 boys,

participated in the evaluation experiment. None of

these students have any experience in dance learning.

Let them learn folk dance and modern dance

successively. These students were divided into an

experimental group and a control group with 23

people in each group. The experimental group entered

a dance class using three-dimensional human

movement analysis to learn dance, while the control

group entered a traditional dance class to learn. In

order to ensure the quality of the experiment, three

judges who have received dance training were invited

to evaluate the dance movements of the students. Only

when at least two people approve a student’s dance

move, is it considered that the student has mastered the

dance move. Finally, the students who mastered the

dance were invited to participate in a questionnaire

survey. Among them, in the study of ethnic dance,

there are 22 people in the experimental group and 19

people in the control group; in the study of modern

dance, there are 23 people in the experimental group

and 22 people in the control group.

3.2.2 Data Sources

This article uses questionnaires to obtain student data,

and analyzes and processes these data to verify the

effect of three-dimensional human movement

analysis on sports dance teaching.

3.2.3 Ways of Identifying

3.2.3.1 Cross-Validation

When learning folk dance and modern dance

successively, a cross-validation method will be used,

that is, the experimental group learning folk dance

will become the control group when learning modern

dance; and the control group learning folk dance will

become the experiment when learning modern dance

group. In this way, all 46 students have the

opportunity to learn dance in two different ways.

3.2.3.2 T-test

When verifying the learning efficiency, because the

learning time is a continuous variable, this article uses

the paired sample test method in the t-test to verify

the efficiency of the three-dimensional human

movement analysis for students’ dance learning. The

paired sample test statistics are:(1),(2),and (3).

𝑡=

𝑑

−μ

𝑠



√

𝑛

⁄

)

𝑑

∑

𝑑





1

𝑛

)

𝑠





∑













1

1

)

Among them, d



is the average of the paired

sample difference, s



is the standard deviation of the

paired sample difference, and n is the number of

paired samples. This statistic t obeys the t distribution

with n−1 degrees of freedom under the condition that

μ=μ0 is true.

3.2.3.3 Chi-Square Test

In the questionnaire survey, because the study is a

non-continuous variable, this paper uses the chi-

square test method to test the help of three-

dimensional human movement analysis to students'

dance learning and the cognitive burden it provides

and the degree of student satisfaction.

4 THREE-DIMENSIONAL

HUMAN MOVEMENT

ANALYSIS AND ITS

EXPERIMENTAL RESEARCH

ANALYSIS IN SPORTS DANCE

TEACHING

4.1 Three-Dimensional Human

Movement Analysis-Error Analysis

The average value of all limb joint coordinate errors

is used as the error of each frame. Table 1 shows the

estimation errors of the human head, pelvis, upper

and lower arms, and upper and lower legs in the first

300 frames.

It can be seen from Figure 1 that when the 3D

human movement analysis analyzes the movements

in the dance process, the estimation errors of the

human head, pelvis, upper and lower arms, and upper

and lower legs in the first 300 frames are all below

70mm. All the above results show that the 3D human

motion analysis performs well in human body posture

tracking, can more accurately analyze the current

INCOFT 2025 - International Conference on Futuristic Technology

230

standardization of dance movements, and can help

students learn dance better.

Table 1: Estimation Error of Part of Human Body Joints

During Dance

Hea

Pelvi

Lowe

r Arm

Uppe

Arm

Calve

Thig

50 22.5 16.4 54.7 37.4 31.4 25.8

25.7 20.1 26.8 26.4 27.3 26.7

22.6 22.4 62.1 32.7 43.6 27.1

36.7 31.6 54.3 34.1 32.2 31.2

34.5 32.4 52.1 33.4 34.6 32.3

39.8 32.7 52.9 55.2 31.7 32.8

Figure 1: Estimation Error of Part of Human Body Joints

During Dance

4.2 An Evaluation Experiment in A

Dance Classroom Using Three-

Dimensional Human Movement

Analysis

Table 2 shows the results of T test and Chi-square test

on the efficiency of the system, the cognitive burden

of the students, the usefulness of the system, and the

degree of satisfaction of the students when the

experimental group and the control group are learning

two types of dance.

Table 2: Test Result

High

Efficien

Cogniti

Burden

Usefulne

Satisfacti

on Level

Nation

Dance

0.0007 0.0008 0.0004 0.0007

Moder

Dance

0.0003 0.0006 0.0008 0.0003

Figure 2: Test Result

From Figure 2, it can be seen that the two groups

of middle school students have significant differences

in the efficiency of the system, the cognitive burden

of the students, the usefulness of the system, and the

average evaluation of satisfaction. The P values of the

test results are all below 0.001. It can be concluded

that dance teaching classrooms using 3D human

movement analysis are more efficient, more helpful,

and more satisfying than traditional teaching

classrooms. The 3D human movement analysis

learning environment can provide a more suitable

cognitive burden.

5 CONCLUSIONS

Nowadays, as the pace of life accelerates, people have

fewer and fewer opportunities to exercise. The

popularization of sports and dance art has become

very difficult, and there are many reasons why people

suffer because of the lack of correct guidance

methods to help them exercise effectively. This

article is dedicated to the development of a sports

dance teaching system based on 3D human movement

analysis. Our goal is to record the coach’s movements

and provide a three-dimensional display effect to help

later students learn the movements better. At the same

time, the system should have the ability to analyze the

quality of sports learning, can point out and correct

the errors of students' local joint points in the process

of sports learning, and can also perform an overall

assessment of the overall sports learning quality of

students.

100

0 50 100 150 200 250 300

Error(mm)

Frame

Head Pelvis

LowerArm Upper Arm

Calves Thigh

0.0005

0.001

0.0015

Test Result(P)

Evaluation Index

National Dance Modern Dance

Analysis of Three-Dimensional Human Movements and Its Research in Sports Dance Teaching

231

ACKNOWLEDGMENTS

Xinjiang University Doctoral Research Funding

Project.(NO:BS190128)

REFERENCES

Purvis, Denise. Dance Science: Anatomy, Movement

Analysis, Conditioning[J]. Journal of Dance Education,

2016, 16(4):157-157.

Wei Z . Application and implementation of motion capture

and motion analysis technology in dance teaching[J].

Revista de la Facultad de Ingenieria, 2017, 32(16):474-

480.

D, Borgogno, A, et al. Test-electron analysis of the

magnetic reconnection topology[J]. Physics of Plasmas,

2017, 24(12):122303-122303.

Hooshang H , Eric T , Li B , et al. Towards a Cybernetic

Model of Human Movement[J]. Mechanical

Engineering Research, 2016, 6(1):29

Yamazaki A , Ikeda T , Tsutsumi T . Main sequence of

torsional saccadic eye movement – analysis by three-

dimensional video-oculography[J]. Acta Oto-

Laryngologica, 2019, 139(1):1-3.

Sun G , Chen W , Li H , et al. A Virtual Reality Dance

Self-learning Framework using Laban Movement

Analysis[J]. JOURNAL OF ENGINEERING

SCIENCE AND TECHNOLOGY REVIEW, 2017,

10(5):25-32.

Wang X , Huang T , Wang W . Human Skeleton Three-

Dimensional Reconstruction-Based Sports Features

Analysis[J]. Journal of Computational & Theoretical

Nanoscience, 2016, 13(12):9988-9992.

Jiang X , Zhang Y , Bai S , et al. Three-dimensional

analysis of craniofacial asymmetry and integrated,

modular organization of human head[J]. International

Journal of Clinical and Experimental Medicine, 2017,

10(8):11424-11431.

Migliorati M , Cevidanes L , Sinfonico G , et al. Three

dimensional movement analysis of maxillary impacted

canine using TADs: a pilot study[J]. Head & Face

Medicine, 2021, 17(1):1-10.

Shogo O , Yasuhiro E , Ryuta S , et al. Three-dimensional

kinematic analysis of glenohumeral, scapular, and

thoracic angles at maximum shoulder external rotation

associated with baseball shadow pitching: comparison

with normal pitching[J]. Journal of Physical Therapy

Science, 2018, 30(7):938-942.

Fan H , Zheng H , Zhao J . Three-dimensional

discontinuous deformation analysis based on strain-

rotation decomposition[J]. Computers and

Geotechnics, 2018, 95(mar.):191-210.

Peng C , Pan B Z , Xue L F , et al. Three-dimensional

structural modeling and deformation analysis of

Archean magnetite quartzite from the Anshan–Benxi

area, northeastern China[J]. Geomechanics and

Geophysics for Geo-Energy and Geo-Resources, 2020,

6(4):62.

INCOFT 2025 - International Conference on Futuristic Technology

232