CULTURAL VISION OF DANCE EDUCATION:

RESEARCH ON TEACHING AND

PERFORMING CHINESE CLASSICAL DANCE

IN CROSS-CULTURAL CONTEXTS

Ke Li*

• Division of Arts, ShenZhen University, Shenzhen, Guangdong, 518060, China

•coco810@sohu.com

Reception: 11 March 2024 | Acceptance: 10 May 2024 | Publication: 11 July 2024

Suggested citation:

Li, K. (2024). Cultural Vision of Dance Education: Research on Teaching

and Performing Chinese Classical Dance in Cross-Cultural Contexts. 3C

Empresa. Investigación y pensamiento crítico, 13(1), 81-100. https://doi.org/

10.17993/3cemp.2024.130153.81-100

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

ABSTRACT

This paper explores the teaching and performance of classical Chinese dance in a

cross-cultural context, and proposes an innovative deep learning-based model for

generating classical Chinese dance. The model effectively combines dance gestures

and audio features, accurately extracts key features from dance and music data, and

transforms these features into internal representations through an encoder, which in

turn generates dance movements that match the music. The introduction of the

attention mechanism allows the model to focus on the important parts of the music

during the generation process, realizing a more harmonious fusion of classical

Chinese dance and music. Finally, the model of teaching ability module, performance

ability module and choreography ability module is carried out to comprehensively

improve students' dance skills and humanistic literacy, and to develop cross-cultural

communication ability and creativity. The results show that this model excels in dance

innovativeness and ability to convey emotions, receiving high scores of 8.9 and 9.3,

respectively. The research results can provide dancers from different cultural

backgrounds with a new perspective for learning and appreciating classical Chinese

dance, and promote the communication and dissemination of world culture.

KEYWORDS

Deep learning; classical Chinese dance; audio features; decoder; attention

mechanism

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

INDEX

ABSTRACT .....................................................................................................................2

KEYWORDS ...................................................................................................................2

1. INTRODUCTION .......................................................................................................4

2. CHINESE CLASSICAL DANCE MUSIC FEATURE EXTRACTION ........................5

2.1. Audio feature extraction .....................................................................................5

2.2. Action feature extraction ....................................................................................6

3. MODULE DESIGN OF GENERATIVE MODELING FOR CLASSICAL CHINESE

DANCE .....................................................................................................................8

3.1. Generative model network structure ..................................................................8

3.2. Classical dance generator and encoder design .................................................9

3.3. Discriminator design for dance movement evaluation ......................................11

3.4. Realistic Chinese Classical Dance Design ......................................................13

4. TEACHING AND PERFORMING INTERCULTURAL CHINESE CLASSICAL DANCE

5. DISCUSSION OF EMPIRICAL EVIDENCE ON THE RESULTS OF TEACHING

CLASSICAL CHINESE DANCE .............................................................................16

5.1. Analysis of Choreographic Authenticity and Coherence ..................................16

5.2. Multidimensional Assessment of Dance Performance .....................................17

6. CONCLUSION ........................................................................................................18

REFERENCES ..............................................................................................................19

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

1. INTRODUCTION

With the deepening of globalization, cultural exchange and integration have

become the theme of the times. In this context, dance, as one of the oldest art forms

of mankind, its education and performance are naturally endowed with a broader

cultural vision [1]. Chinese classical dance has a long history and rich movements,

and is an important part of Chinese culture. With its unique rhythms, gestures and

expressions, dance shows the charm of Chinese civilization to the world [2-3].

However, due to cultural differences and language barriers, the teaching and

performance of classical Chinese dance in cross-cultural contexts often face many

challenges. In terms of teaching, due to the differences in students' perception and

understanding of body language in different cultures, how to enable students to

accurately master the movement essentials and rhythmic characteristics of classical

Chinese dance has become the primary problem that teachers need to solve [4]. In

terms of performance, it is also a big challenge to make audiences from different

cultural backgrounds understand and appreciate the beauty of classical Chinese

dance. Teachers and performers are required to have not only superior dance skills,

but also profound cultural literacy and cross-cultural communication skills [5].

Combined with the research data, it is found that a large number of scholars put

forward targeted strategies for the development of classical Chinese dance and

foreign dance teaching and performance. For example, Ma, L perfectly united training,

creation and performance in the teaching of classical Chinese dance, analyzed the

importance and artistic characteristics of body rhyme teaching, explored the

comprehensive scoring system from four aspects, namely, the establishment of

evaluation index dimensions, the analysis and treatment of evaluation indexes, the

weight ratio of evaluation index dimensions, and the conversion and comprehensive

calculation of evaluation indexes, and explored the new possibilities of choreographic

creation [6]. Li, J et al. on the MOOCs platform, integrating high-quality teaching

resources worldwide, including dance videos, teaching courseware, expert lectures

and so on. And using the MOOCs platform's wide dissemination and sharing function,

the customized teaching method can deliver the quality dance teaching resources to

more people and promote the inheritance and development of the dance art [7]. Yang,

X in the special performance prediction model for dance students, correlation analysis

and clustering analysis were performed, and neural network was used to build the

prediction model, and the model was analyzed for reliability and error. The test results

showed that the model constructed by the neural network predicted the level of

development of dance performance and physical fitness more accurately than the

regression model [8]. Anderson, M presents the work of many of Australia's major

dance companies and individual performers while spanning a wide range of genres,

including modern, ballet, theater, contemporary, folk and social dance styles. This

dance collection is part of the Australian Performing Arts Collection and encompasses

five key areas of circus, dance, opera, music and theatre [9]. Zhang, D takes dance

choreographers as the object of study and applies the music feature extraction

technique in machine learning to analyze the choreographers, applying the extracted

music feature information to the choreography process while synchronizing the audio

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

and video to achieve better choreographic effect [10]. Daly, D. K adopted an artistic

practice investigation approach that included designing, rehearsing, performing and

recording two performance events. A departure from traditional classical music

methods, drawing on Dalcroze's artistic rhythmic techniques and utilizing ethnography

and other arts-based methods to develop and collect data. An overview of Dalcroze's

artistic rhythms from the performer's perspective, i.e. the researcher's, is presented

and the implications of this approach for the investigation of his artistic practice are

discussed [11].

The above dance teaching methods are often limited to a particular cultural context,

which is difficult to meet the growing demand for cross-cultural teaching. In order to

break through this limitation, this paper starts from the cultural perspective of dance

education, and the deep learning-based machine dance automatic generation

technology brings new possibilities for cross-cultural dance teaching. Deep learning

technology's has achieved remarkable results in the fields of image recognition,

speech recognition and natural language processing. Deep learning has a powerful

feature extraction and learning ability, which can automatically extract useful

information from a large amount of data and simulate the complex thinking process of

human beings. Therefore, applying deep learning technology to the generation and

teaching of Chinese classical dance is expected to provide a new solution for cross-

cultural dance communication and dissemination. Through deep learning algorithms,

a classical Chinese dance generation model is constructed, which is able to combine

dance gestures and audio features, improve the model's ability to capture the complex

relationship between dance and music by improving the traditional encoder and

decoder and introducing the attention mechanism, so as to make a perfect match

between dance and music, realize a better fit between classical Chinese dance

movements and music, and provide a new perspective for dancers of different cultural

backgrounds to learn This will provide dancers from different cultural backgrounds

with a new perspective on learning and appreciating classical Chinese dance. At the

same time, by combining the teaching ability module, performance ability module and

choreography ability module, we can comprehensively improve students' dance skills

and humanistic qualities, and cultivate cross-cultural communication ability and

creativity.

2. CHINESE CLASSICAL DANCE MUSIC FEATURE

EXTRACTION

2.1. AUDIO FEATURE EXTRACTION

Audio feature extraction is a key technique for analyzing and understanding the

intrinsic properties of music, especially for Chinese classical dance music, whose

unique rhythmic and rhythmic characteristics need to be captured by precise audio

analysis. This process not only involves the technical aspects of signal processing,

but also requires an in-depth understanding of the cultural connotation and artistic

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

expression of the music. In terms of rhythmic feature extraction, Mel Frequency

Cepstrum Coefficient (MFCC) is a commonly used method. MFCC is able to simulate

the human ear's perceptual characteristics of sound and effectively represent key

information such as pitch, timbre, etc. of an audio signal. By extracting MFCC

features, the melody of classical Chinese dance music can be analyzed in detail to

capture its unique pitch changes and timbre characteristics. Rhythm feature

extraction, on the other hand, is an important part of understanding classical Chinese

dance music. Classical Chinese dance music often has a distinctive sense of rhythm,

which is crucial to the coordination of dance movements. Rhythm feature extraction

needs to accurately capture the beat information in the music, including the strength

and speed of the beat. By analyzing these rhythmic features, the coordinated

relationship between music and dance movements can be better understood. In

conclusion, through the extraction of rhythmic and rhythmic features, the intrinsic

properties of music can be understood more deeply, providing strong support for the

subsequent generation and teaching of dance movements.

2.2. ACTION FEATURE EXTRACTION

Action feature extraction is a key link to deeply understand the connotation and

style of dance movement, especially in the analysis of Chinese classical dance, which

is famous for its rich body language, precise action positioning and smooth action

transition, so extracting its action features is of great significance for subsequent

dance generation, teaching and performance. OpenPose technology can detect and

track multiple key points of the human body in real time. The OpenPose technology

can detect and track multiple key points of the human body in real time, and Figure 1

shows the key points of human dance movement characteristics. Figure 1 shows the

key points of human dance movements, such as wrists, ankles, knees, shoulders,

etc., so as to accurately capture the position information, trajectory, speed and

acceleration of the key points of dance movements. These features can not only

describe the basic form of dance movements, but also reflect the dancer's

performance style and emotional expression. In Chinese classical dance, the soft

curves of the arms and wrists, the light jumps of the feet, and the coordinated rotation

of the body are important parts of the dance flavor and style, and the 18 key points of

the human body are obtained by extracting the human body movement features using

OpenPose.

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

Figure 1. Key point detection of human dance motion features

By extracting audio features and action features, the following sets of feature data

are obtained:

1. Audio feature

, i.e., the feature dimension of each

audio is 32.

2. Beat feature

, i.e., the beat dimension size of the video

frame section is 3.

3. Pose feature

, i.e., the size of the pose dimension in a

video frame is 36.

4. Audio data , which means that the audio data is

frames in total, and each frame is represented as .

Mi=⟨m1

i,m2

i,⋯,m32

i⟩

Bi=⟨b1

i,b2

i,⋯,b3

i⟩

pi=⟨p1

i,p2

i,⋯,p36

i⟩

M={M1,M2,⋯,Mn}

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

5. Dance Gesture Data , indicates that the dance gesture is

frames, and each frame is represented as .

3. MODULE DESIGN OF GENERATIVE MODELING FOR

CLASSICAL CHINESE DANCE

3.1. GENERATIVE MODEL NETWORK STRUCTURE

After feature extraction is completed, the classical Chinese dance generation

model can be divided into three main parts, generator, self-encoder and discriminator.

These parts work together to generate realistic dance movements. The network

structure of the classical Chinese dance generation model is shown in Figure 2. The

generator is the core part of the dance generation model, which is responsible for

generating the corresponding dance action gestures based on the input audio

features or other information. Self-encoder is a neural network model for data

compression and reconstruction, which is used for audio reconstruction in the

classical Chinese dance generation model. The self-encoder accepts the input audio

signal and tries to recover the original audio signal through the process of encoding

and decoding [12]. In the dance generation model, the role of the self-encoder is to

compress and reconstruct the extracted audio features so that they can better guide

the generator to output the corresponding dance movements. The discriminator is

another important component used to evaluate the authenticity of the generated

classical Chinese dance movements. The task of the discriminator is to determine

whether the generated classical Chinese dance movements are similar to the real

ones, so as to help the generator learn more realistic dance movements. By

combining the three parts of generator, self-encoder and discriminator, a complete

model of classical Chinese dance generation can be constructed. This model is able

to generate realistic classical Chinese dance movement gestures based on input

audio features or other information, which provides powerful support for classical

Chinese dance creation and learning [13-14].

p={p1,p2,⋯,pn}

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

Figure 2. Network structure of Chinese classical dance generation model

3.2. CLASSICAL DANCE GENERATOR AND ENCODER DESIGN

In order to make the generation of Chinese classical dance more effective, the

Seq2Seq model is chosen to design the generator, which specifically includes the

encoder and decoder. However, the Seq2Seq model cannot represent all the

semantic information, and the attention mechanism will be added on the basis of this

model, so as to become a context vector that can summarize different semantic

information at different moments. The dance generation sequence with the addition of

the attention mechanism is shown in Figure 3.

In the encoder part, the attention mechanism is used to calculate the weight of

each position in the input sequence so that all relevant information is taken into

account when generating the context vector. This is achieved by assigning a weight to

each position in the input sequence, which is calculated based on the importance of

the semantic information at that position. In the decoder part, the same attention

mechanism is used to compute the weight for each position in the output sequence. In

this way, the model can take into account contextual information when generating

each output step and emphasize or ignore certain parts as needed. By combining the

Seq2Seq model and the attention mechanism, a flexible dance generation model can

be created. The model can generate richer, more realistic and compliant dance

movements based on input audio features and other information.

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

Figure 3. Attention mechanism

With the introduction of the attention mechanism, the generator network is able to

focus on different semantic information at different moments to better understand and

generate dance movements related to the input audio features [15]. This design

approach can improve the flexibility and generation effect of the model, and better

meet the needs of cross-cultural teaching and learning. The generator model of

classical Chinese dance for the generator network after the introduction of the

attention mechanism is shown in Figure 4.

Figure 4. Schematic diagram of the Chinese classical dance generator model

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

3.3. DISCRIMINATOR DESIGN FOR DANCE MOVEMENT

EVALUATION

In order to make the matching of classical Chinese dance gestures and music more

accurate, the matching results are evaluated by adding a discriminator. The

discriminator is designed to evaluate how well the generated dance movements

match the music. By adding the discriminator, the generated dance movements can

be evaluated more accurately so that the generation process can be adjusted to

improve the matching between the dance movements and the music. The

discriminator receives the generated dance movement as input and outputs an

evaluation result indicating how well the movement matches the music. This

evaluation result may be a probability value, e.g., the discriminator outputs a

probability value close to 1 if the generated dance movement is highly matched to the

music, and a probability value close to 0 if the degree of matching is low. During the

training process, the loss function of the discriminator can be designed according to

practical needs. A common loss function is the binary cross-entropy loss, which is

used to measure the difference between the generated movements and the real

music match. By minimizing this loss function, the discriminator can gradually learn

how to accurately assess the degree of match between dance movements and music.

The specific way of discriminating is mainly through the loss function of the

(1)

Where, denotes the generator, denotes the music,

denotes the

generation of the dance gesture, and denotes the size of the probability that

the dance matches the music, if the value is close to 1, it indicates that the two match

more. If the value is in the range of 0, it indicates that the dance-music matching

probability is small.

The classical Chinese dance gesture encoder network structure is shown in Figure

5, after generating the dance gesture, the discriminator is used to judge the match

between the dance and the audio, so as to confirm whether the audio is consistent

with the real gesture. If the gesture vector is denoted as

, the

difference between the front and back frames is denoted as

, and the audio features are denoted as

[16]. All of them are input into the discriminator and then the

discrimination is calculated by equation (1).

GA N(G,D)=E(P,M)[log D(P,M)] +

[log(1 −D(P,M)) + log(1 −D(W,M))

]

G(M)

D(G,M)

p={p1,p2,⋯,pn}

{P1−P2,P2−P3,⋯,Pn−1−Pn}

M={M1,M2,⋯,Mn}

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

Figure 5. Network structure of Chinese classical dance pose encoder

The role of self encoder is to reconstruct the audio features. If

denotes an audio

feature and

denotes a beat feature, the match between music and dance can be

improved by inputting the two features into the encoder and then encoding and

decoding them. The specific reconstruction is as follows:

(2)

(3)

where denotes the reconstructed audio,

denotes the extracted low

dimensional audio features.

denotes the splicing process of the extracted

parameters, and and

denote the neural networks to be learned

[17].

fi=Encoder(Concat(Mi,Bi))

M∼

i=Decoder(fi)

M∼

Concat

Encoder

Decoder

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

3.4. REALISTIC CHINESE CLASSICAL DANCE DESIGN

In order to make the generated dance movements more realistic and in line with

the style of Chinese classical dance, the improved Pix2Pix algorithm can be used to

further process the generated dance movements. The training process of the

improved Pix2Pix network for the dance movements is shown in Figure 6. The Pix2Pix

algorithm is an image-to-image conversion algorithm that converts the sketches of the

dance poses into real dance poses. During the training process, the Pix2Pix algorithm

gradually improves its conversion ability by learning the mapping relationship from

sketches to real dance poses. The algorithm continuously optimizes its parameters to

minimize the difference between the real dance pose and the conversion result. By

using the improved Pix2Pix algorithm, the generated dance movements can be

converted into movements that are more realistic and in line with the Chinese

classical dance style.

Figure 6. Dance Action Training Process of Improved Pix2Pix Network

Based on the above training process, the whole process of realizing the live-action

dance can be divided into the following steps:

Video reading and single-frame extraction, using the OpenCV library to read

the dance video file. Extract single frames in the video to form a collection of

single frames as input for subsequent processing.

Dance movement pose extraction, use pose estimation tools such as

OpenPose to analyze the extracted single frames for pose analysis. The pose

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

information is converted into a heat map representation, where each pixel of

the heat map represents the action pose of the corresponding part.

Establish the mapping relationship between the heat map and the real-life

video, using the labeled training dataset, train the Pix2Pix model to establish

the mapping relationship from the heat map to the real-life dance video. Let the

model learn how to convert the pose information into real dance movements.

Dance movement coordinate conversion and video synthesis to convert the

predicted dance gesture coordinates into corresponding heat maps. Using the

trained Pix2Pix model, the newly generated heat maps are converted back to

real dance movements. These real dance action frames are combined

according to the chronological order to form a new real-life version of classical

Chinese dance video, which further enriches and advances the teaching of

classical Chinese dance.

4. TEACHING AND PERFORMING INTERCULTURAL

CHINESE CLASSICAL DANCE

In order to further enrich and promote the teaching and dissemination of cross-

cultural classical Chinese dance, the cultivation model of teaching ability module +

performance ability module + choreography ability module can be implemented. This

model aims to cultivate students' intercultural communication ability and dance ability

through systematic teaching and practicing activities. The three levels of modules for

the cultivation of classical Chinese dance ability are shown in Figure 7.

The basic skills training in the teaching ability module is the cornerstone of

dance learning, including the training of body flexibility, strength and

coordination. Body rhythm is the soul of dance, emphasizing the integration of

the dancer's inner emotions and outer movements [18]. Understanding the

development history of classical Chinese dance as well as teaching methods

provides future dance teachers with the necessary knowledge base. By

attending and observing various performances, students can cultivate

creativity, stimulate vigor and passion, and better understand the charm of

dance art.

The performance ability module includes repertoire rehearsals and

international exchange activities. By rehearsing classical or modern Chinese

classical dance repertoires, students can improve their performance skills and

stage presence. Students are encouraged to participate in international dance

festivals, cultural exchange activities, etc., to cultivate their international vision

and cross-cultural communication skills [19-20].

The choreography ability module includes the fundamentals of dance

choreography, multicultural integration, and international cooperation

programs. Learning the basic principles and techniques of choreography

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

enables students to create dance works with personal characteristics [21].

Students are guided to explore the possibility of combining classical Chinese

dance with dance elements from other countries and regions to develop their

cross-cultural choreographic ability. Encourage students to participate in

international dance creation projects and collaborate with international dance

artists to create dance works with international perspectives [22].

This cultivation mode of curriculum module + practical ability adds the element of

cross-cultural exchange to the original foundation, provides students with a wider

range of dance education opportunities, and cultivates their ability to perform, teach

and create dance in the context of globalization. Through participation in international

exchange activities and cross-cultural cooperation programs, students can better

understand and disseminate Chinese classical dance culture, and at the same time

learn from and absorb the outstanding dance elements of other countries and regions,

thus laying a solid foundation for their future cross-cultural teaching and performing

careers.

Figure 7. Three Levels of Cultivating the Ability of Chinese Classical Dance

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

5. DISCUSSION OF EMPIRICAL EVIDENCE ON THE

RESULTS OF TEACHING CLASSICAL CHINESE

DANCE

5.1. ANALYSIS OF CHOREOGRAPHIC AUTHENTICITY AND

COHERENCE

In order to objectively assess the authenticity and coherence performance of the

system proposed in this paper on classical Chinese dance choreography, a user

manual scoring method was adopted for the empirical study. Thirty-five postgraduate

students were invited as experimental participants, 18 of whom had a background in

dance, musical instruments or voice. This combination of participants ensured the

professionalism of the scoring and also took into account the aesthetic viewpoints of

the general audience. During the experiment, emphasis was placed on ensuring that

the participants had sufficient musicality and aesthetic knowledge of music and dance

to ensure the reliability and validity of their ratings. Participants were asked to watch

two videos of classical Chinese dances, one was a classical Chinese dance

choreographed through traditional teaching methods, and the other was a classical

Chinese dance generated through the methodology of this paper. After viewing,

participants were asked to rate the authenticity and coherence of the two dances on a

five-point scale based on their subjective judgment. The scale was 1-5, where 1

indicated very poor and 5 indicated very good. This is designed to allow participants to

express their evaluation clearly. The result scores for the authenticity and coherence

of the choreography of classical Chinese dance are shown in Figure 8, and the

authenticity scores of the dance movements generated by this paper's method are all

significantly higher than the results of the control group.

Figure 8(a) shows the authenticity score, the final score of the authenticity of the

choreography results of the control group is 1.61, and the proportion of users who

scored 1 and 2 is more than 40%, i.e., most of the users think that the results of the

choreography are very inauthentic and inauthentic, and the users who think that the

results are barely authentic and authentic only account for 6% each. The final score

for the authenticity of the results of this paper's method is 4.09, with more than 60% of

users scoring 4 and considering the results to be authentic, and 24% of users

considering the results to be very authentic. Figure 8(b) shows the coherence score,

the coherence of the results of the control group choreography final score of 1.76, the

highest score is only 3, the users who scored 1 and 2 almost accounted for 50% each,

i.e., the users who think the results are very incoherent and incoherent almost

accounted for half each. The final score for the coherence of the results of this paper's

method was 4.1, with a minimum score of 3. More than 70% of users scored 4 and

considered the results coherent, and almost 20% scored 5 and considered the results

very coherent.

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

Figure 8. The choreography results of Chinese classical dance

5.2. MULTIDIMENSIONAL ASSESSMENT OF DANCE

PERFORMANCE

In order to comprehensively assess the effectiveness of the dance auto-generation

method in teaching classical Chinese dance, an empirical study was designed for

dance performance. Two groups of experimental subjects with different backgrounds5

were selected for this study, aiming to compare the differences between traditional

teaching methods and supplementary teaching methods incorporating dance auto-

generation in developing students' dance performance abilities. Two groups of

experimental subjects were selected, one was students with a basic knowledge of

classical Chinese dance, experimental group A, and the other was students without a

basic knowledge of dance but with an interest in Chinese culture, experimental group

B. A four-week cross-cultural classical Chinese dance teaching program was

conducted for the two groups of students, in which experimental group A adopted the

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

traditional teaching method, and experimental group B used the dance auto-

generation system as the supplemental teaching method.

At the end of the teaching period, students in both groups were assessed on their

performances, and the results of the multidimensional classical Chinese dance

performance assessment are shown in Figure 9. In all three dance actions, the scores

of experimental group B were higher than those of experimental group A.

Experimental group A was good in the accuracy of dance actions, and scored 8.5

points in dance action 1. It scored relatively low on rhythm, expressiveness, creativity,

and ability to convey emotion. Experimental group B scored higher on all assessment

dimensions than experimental group A. It performed well in dance action 3 with a

score of 9.6 for movement accuracy. There were also significant improvements in

rhythm, expressiveness, innovativeness, and the ability to convey emotion, especially

in innovativeness and the ability to convey emotion, which received high scores of 8.9

and 9.3, respectively. From the data, experimental group B combined with the dance

automatic generation system teaching showed a trend of superiority over the

traditional teaching method of experimental group A in all aspects of dance

performance. This suggests that the use of the Dance Automated Generation System

in teaching may help to improve students' dance performance abilities, especially in

terms of creativity and ability to convey emotion. These enhancements may be

attributed to the features of instant feedback, personalized instruction, and creative

inspiration provided by the system.

Figure 9. Evaluation Results of Multidimensional Chinese Classical Dance Performance

6. CONCLUSION

In order to promote the inheritance and promotion of Chinese classical dance, this

paper successfully constructs a deep learning-based Chinese classical dance

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

generation model, which realizes a better fit between dance movements and music by

combining dance gestures and audio features. The results of the traditional teaching

method and the teaching method combined with the automatic dance generation

system on the assessment of students' performances were compared. The final score

of the authenticity of the choreography results of the control group was 1.61, and most

users considered the choreography results to be very inauthentic and unrealistic. The

final score for the veracity of this paper's method was 4.09, with more than 60% of

users considering the results to be veritable and 24% considering the results to be

very veritable. The final score for the coherence of the choreography results for the

control group was 1.76, and most of the users considered the results to be very

disjointed and incoherent. The final score for the coherence of this paper's method

was 4.1, with more than 70% of users considering the results to be coherent and

nearly 20% considering the results to be very coherent. Experimental Group B, taught

in conjunction with the Dance Automation Generation System, showed a tendency to

outperform the traditional teaching methods of Experimental Group A in all aspects of

dance performance. The cultural vision of dance education in the study of teaching

and performing classical Chinese dance in a cross-cultural context has both deep

cultural connotations and is able to promote cultural inheritance and innovation, as

well as the coexistence and development of global cultural diversity.

REFERENCES

(1) Flønes, M. (2023). Diffracting through an ethical itch: Becoming a response-able dance

education researcher. Journal for Research in Arts and Sports Education, 7(2).

(2) Mabingo, A. (2022). Decolonizing assessment in dance education: Ubuntu as an

evaluative framework in Indigenous African dance education practices. Journal of Dance

Education, 22(1), 1-12.

(3) Yeum, H. J., & You, J. Z. (2022). A study of new direction of Korean dance education

according to digital conversion policy. Journal of Convergence Consilience.

(4) Nam, J. (2020). Formation and diversification of contemporary Chinese classic dance.

The Eastern Classic Studies, 42, 241-270.

(5) Kasmahidayat, Y., Hanafiah, M., & Hasanuddin. (2022). Gending Sriwijaya dance

performance as an innovation media to develop the cross-culture based local economy.

Advances in Social Science, Education and Humanities Research.

(6) Ma, L. (2022). Operation design of comprehensive scoring system for classical dance

training based on information technology. 2022 International Conference on Information

System, Computing and Educational Technology (ICISCET), 4-6.

(7) Li, J., Kou, H., Wang, J., & Ren, W. (2023). Enhancing online learning for dance majors:

A customized teaching approach using massive open online courses. Education and

Information Technologies, 1-29.

(8) Yang, X. (2022). Analysis of the construction of dance teaching system based on digital

media technology. Journal of Interconnection Networks, 22(Supp05), 2147021.

(9) Anderson, M. (2020). Dance overview of the Australian Performing Arts Collection.

Dance Research, 38(2), 149-167.

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

(10) Zhang, D. (2022). Research and implementation of choreographer technique based on

music feature extraction. Journal of Interconnection Networks, 22(Supp02), 2143036.

(11) Daly, D. K. (2022). Creativity, autonomy and Dalcroze Eurhythmics: An arts practice

exploration. International Journal of Music Education, 40(1), 105-117.

(12) Bi, J., Wang, Z., Yuan, H., Zhang, J., & Zhou, M. (2023). Self-adaptive teaching-learning-

based optimizer with improved RBF and sparse autoencoder for high-dimensional

problems. Information Sciences, 630, 463-481.

(13) Hilmkil, A., Thomé, C., & Arpteg, A. (2020). Perceiving music quality with GANs. arXiv

preprint arXiv:2006.06287.

(14) Ma, X., Wang, Y., Kan, M. Y., & Lee, W. S. (2021, October). AI-lyricist: Generating music

and vocabulary constrained lyrics. In Proceedings of the 29th ACM International

Conference on Multimedia (pp. 1002-1011).

(15) Wang, W. Y., Du, W. W., & Peng, W. C. (2022). RecFormer: Personalized temporal-

aware transformer for fair music recommendation. In CEUR Workshop Proceedings (Vol.

3318). CEUR-WS.

(16) Qin, Y., Huang, T., & Tang, G. (2022). A hierarchical children’s dance movement pose

estimation method based on sequence multiscale feature fusion representation.

Advances in Multimedia, 2022.

(17) Wang, S., Li, J., Cao, T., Wang, H., Tu, P., & Li, Y. (2020). Dance emotion recognition

based on Laban motion analysis using convolutional neural network and long short-term

memory. IEEE Access, 8, 124928-124938.

(18) Amie, K. (2022). Hidden curriculums of the black dancing body: Teaching within and

among racialized perceptions. AERA 2022.

(19) Park, S. (2023). A study on importance and satisfaction of attending a dance

performance using importance-performance analysis (IPA). Journal of the Korea

Entertainment Industry Association.

(20) Ren, H. (2023). Modern strategies for dance teacher performance management at dance

colleges on the example of Central China. Research in Dance Education, 1-15.

(21) Bilitza, M. S. (2021). Being the facilitator: A brief research report on the motivation of the

choreographer and dance maker to work with heterogeneous groups in a community

dance setting. Frontiers in Psychology, 12, 601033.

(22) Hodgson, A. (2020). Beyond the Opera House: Some Victorian Ballet Burlesques. Dance

Research, 38(1), 7-24.

https://doi.org/10.17993/3cemp.2024.130153.81-100

3C Empresa. Investigación y pensamiento crítico. ISSN: 2254-3376

Ed. 53 Iss.13 N.1 January - March, 2024

100