Intelligent Performance Evaluation in Dance Training

 

Eeshita Goyal ¹, Likhith S R ², Girish Kalele ³, Divya Sharma ⁴, Santosh Ku ⁵, Shreyas Dingankar ⁶

 

¹ Assistant Professor, School of Business Management, Noida International University 203201, India

² Assistant Professor, Department of Computer Science and Engineering, Presidency University, Bangalore, Karnataka, India

³ Centre of Research Impact and Outcome, Chitkara University, Rajpura- 140417, Punjab, India

⁴ Chitkara Centre for Research and Development, Chitkara University, Himachal Pradesh, Solan, 174103, India

⁵ Assistant Professor, Centre for Artificial Intelligence and Machine Learning, Institute of Technical Education and Research, Siksha 'O' Anusandhan (Deemed to be University) Bhubaneswar, Odisha, India

⁶ Assistant Professor, Institute of Management and Entrepreneurship Development Pune, Bharati Vidyapeeth Deemed to be University Pune, India

 

1. INTRODUCTION

Dance is an art form that is multidimensional and contains combination of physical precision, expression of emotion, rhythms synchronization and identity to the culture. Dance, as an artistic activity and as a technical practice of performance, needs to be constantly fined through the quality of movements, the correctness of poses, balance, time, and expressionism. Conventionally, the measurement of performance in dance was conducted by professional instructors, who base their judgment in the experience and years of expertise, observation and subjective reading. Although it is invaluable, this human-based assessment system is always constrained in the form of bias of perception, diversity among instructors, and difficulty in the provision of real-time, data-driven feedback Li et al. (2021). In this regard, the development of intelligent performance evaluation systems can be considered as a revolutionary practice in the sphere of modern dance training. The convergence of artificial intelligence, computer vision, machine learning, and motion-sensing technologies has made this a new paradigm in which the performance of dancers could be precisely analyzed as never before. Pose estimation systems based on AI like OpenPose, Media Pipe, and neural skeleton trackers can identify the coordinate, angle, trajectory, and spatial pattern of dance moves on the joints Tsuchida et al. (2019). These calculation capabilities allow quantifying such core performance parameters as the accuracy, symmetry, rhythm compliance, weight transfer, and expressive dynamics. Consequently, smart evaluation systems have the ability to objectively evaluate the technical features that might not be easily observed by the human eye, particularly in the challenging and fast choreography Lin (2015).

                                                                         Figure 1

 

The intelligent systems provide audience with continuous and personalized feedback that develops with the development of the dancer. Rather than apply generalized corrections, AI-based models can give up details about recurring errors like delayed footwork, insufficient arm extension, or the lack of consistency in center-of-mass control and suggest the specific micro-corrections. Such accuracy is helpful to dancers of all abilities, including those trying to learn the basic technique and the professionals developing advanced movement vocabulary Zhang (2023). The transition between the occasional feedback of the instructors to continuous data-driven feedback promotes faster skill development and improves the efficiency of training. The latest developments in affective computing and movement emotion recognition have provided the avenues of studying expressive qualities in dance. The patterns of movement-effort, facial expressions, and the distribution of energy through time can be trained with models to assess the elements of emotional clarity, stylistic authenticity that are essential to classical, folk, and contemporary dance styles Zhenyu (2025). This will fill the gap between technique and artistry so that AI systems will not interfere with the expressive integrity that constitutes dance as an art form as shown in Figure 1. Notably, smart performance assessment is also effective in improving safety and injury prevention. Through the joint stress, movement asymmetry, and repetitive strain indicators, AI would be able to warn dancers of high-risk patterns before they are injured. This preventive measure is an aid to the well-being of dancers and their long-term sustainability. Although technological integration has massive benefits, there are such issues as cultural sensitivity, bias in algorithms, and the maintenance of individuality in art. Dance is very cultural, tradition-based and subjective; hence, smart systems should be created to enhance but not to override human creativity. It is necessary to ensure the authenticity of dance traditions with the help of ethical data collection, culturally diverse training data, and open assessment models.

 

 

2. Dance as an Embodied Creative Practice

Dance is essentially an embodied art since it is the body that is the medium and instrument of communication. Dance is not based on any external object like paint, clay, or musical instruments as is the case with other creative practices. This imbibition renders dance distinctly immersive, in that there is creation of meaning through movement, gesture, rhythm, breath, and expressional motive. Bodies of the dancer are the dynamic interface between the internal feelings and outside surroundings and provide an endless negotiation between technique, creativity, and personal interpretation Ishii et al. (2024).

Fundamentally, dance is influenced by kinesthetic awareness the knowledge of the dancer about posture, balance, flow of energy, and space orientation. Training enables the dancers to have an intuitive mastery of muscle coordination, movement directions, and expressive dynamics. Kinesthetic intelligence gives the performers the ability to perform complicated choreographic patterns as well as express emotion and narrative by making slight changes in the quality of movement. Such complicated combination of physical and emotional intelligence makes dance a process in which it is impossible to separate expression and the body itself Chéron et al. (2015).

                                                                         Figure 2

 

Dance is a place of tradition, identity and collective memory, which is culturally bound. Each culture has its set of moving words, gestures, rhythms, and aesthetic values. In the classical tradition of Bharatanatyam, Ballet or Flamenco, mastery of the disciplined movements and the style of nuances is required. Community stories, rituals, and local identity are reflected in folk dances and experimentation, fluidness and individualistic expression adopted in contemporary dance Yao et al. (2011). This heterogeneity underlines the fact that dance is not simply physical movement but is rather a textualized culture of generations of practice, history as well as being a meaning-making process as shown in Figure 2. Emotion is the other important aspect of embodied dance practice. Emotional intention of the dancer has an impact on the movement texture, timing, modulation of energy and the clarity of the expressive value as a whole. Dancers use the use of embodied emotion in order to communicate with their audiences whether it is pleasure or pain, fidelity or defiance Liu et al. (2023).

Dance improvisation, exploration of choreography and individualized interpretation creates creativity in the field of dance. Dancers in even the most rigid classical traditions apply individuality in their performance by micro-differences in expression, timing, and energy. This artistic independence is very fundamental to the artistic character of dance; this guarantees performances to be one in a thousand and highly intimate. Due to this reason, dance cannot be boiled down only to parameters of measures in movement; it should be perceived as a holistic embodied practice that incorporates both technique and human craftsmanship. In terms of training, dancers go through repetitive patterns of practice, contemplation, and perfecting Singh et al. (2021). Embodied learning is dependent on sensory feedback in terms of visual, proprioceptive, auditory, and tactile information that is used to make dancers internalize movement pathways and perfect emotional expression. In that regard, the intelligent evaluation systems should not disregard the embodied character of dance and make sure the computational measures will complement artistic individuality but not limit it. To really help dancers, the smart systems should realize that dance is not only related to performing the right movements but being in the movements with the authenticity of expression.

 

3. Digital Transformation in Performing Arts

Rapid changes in digital technologies have brought the performing arts environment to a major metamorphosis in the last decades. Computational tools, immersive media, motion-sensing systems, and artificial intelligence have had an immensely strong impact on dance, in particular. The technologies have not only transformed the way performances are created, taught and experienced but also the way dancers conceptualize movement, space and embodiment Khosla et al. (2021). With the ever-increasing effects of digital innovation, which is shifting the physical and virtual to become one and the same, dance is presented as an unusually fluid locus of interdisciplinary creative activity as artistic practice and technological intelligence intersect. The transformation of digital motion capture systems can be listed among the most influential processes. Motion capture was first utilized in animation and games but now it is an effective instrument to be used by choreographers and teachers who need to analyze body movements in a very specific way. Optical markers, inertial sensors, depth cameras (e.g., Kinetc, LiDAR) can enable choreographers to see the trajectories, timing patterns, spatial formations and qualities of movement in the three-dimensional space Ullah et al. (2022). The tools allow the analysis on a level that was hitherto not able to be seen with the human eye, which allows the choreographer to make more informed choices and provides the dancer with a comprehensive grasp of their movement structure.

                                                                         Figure 3

 

Parallel to motion capture, interactive dance training has had new possibilities due to the rise of immersive technology including virtual reality (VR), augmented reality (AR) and mixed reality (MR). VR technologies enable dancers to interact with virtual teachers, practice in virtual theatres, and preview the intricate compositions in different perspectives. AR superimposes virtual instructions on real-life practice sessions and helps dancers correct their poses, set time, and position in space as in Figure 3. The use of these immersive tools supplements embodied learning and the result is a more interactive, approachable and interesting training experience among dancers of different skill levels. Artificial intelligence has been a major factor in defining the new digital dance practices. Models based on machine learning trained on datasets representing dance styles of various types can be used to create original movement sequences and help dancers in their creative work and experimentation as well as to facilitate their improvisation of dance tasks Lee et al. (2022). Such a combination of AI creates the partnership between human creativity and computational intelligence, pushing the limits of what it takes to be choreographic authorship. Furthermore, there has been the digital transformation of the spread and preservation of dance. Digital libraries, Web-based platforms, and motion databases make a difference in making dance traditions accessible to people worldwide and thus safeguarding the continuation of culturally relevant traditions. Digital media can record nuances of style, gesture, rhythm, and emotionality, and transmit it to future generations as well as facilitate cross-cultural exchange. Remote learning and AI-driven system of instruction will enable students in distant areas to receive world-general training, fostering inclusivity and making the performing arts democratic Takahashi et al. (2018)

 

4.  Computational Representation of Aesthetics

The aesthetics of dance is a complicated combination of technology, expression, rhythm, space, and cultural significance. One of the most difficult and, at the same time, the most transformative features of intelligent dance evaluation is the translation of these most human and subjective qualities into computationally measurable parameters. Computational aesthetics attempts to encode artistic features in form of quantifiable data to enable that machine learning algorithms can perceive the quality of movement, the intent of style and its emotional impact without compromising the aesthetic quality of the performance. Aesthetic computation is based on estimate computation of poses, as keypoints of skeletal keypoints that are joint locations and limb orientations. These keypoints are the units upon which the higher-level features of joint angles, movement trajectory, symmetry patterns and center-of-mass dynamics are derived Malleson et al. (2020). Nevertheless, pose estimation is a problem that can only capture structural accuracy whereas aesthetics require more subtle descriptors. Consequently, other layers of analysis are based on the time-related features like rhythmic alignment, transition smoothness, velocity variation and phrasing of movement patterns. These time measures give an estimation of the musicality, coordination and the flow of expression by the dancer.

Table 1

Table 1 Dataset Types for Computational Aesthetics Modeling
Dataset Type	Contents	Usage in Dance Evaluation	Challenges
Skeleton Keypoint Data	Joint coordinates	Technique analysis	Loss of expressiveness
Video Datasets	Full body + facial cues	Style and emotion detection	High storage
Motion Capture Files	3D trajectories	Precision evaluation	Equipment constraints
Audio–Movement Pairs	Music + movement	Musicality assessment	Synchronization issues

 

Along with structural and temporal parameters, there is aesthetic that involves stylistic identity. Computerized models apply feature engineering to encode style-specific qualities to the instances such as roundedness and footwork complexity in Bharatanatyam, fluid motions in the torso as part of contemporary dance or even sharp flamenco rhythmic movements. Style classification models are models trained on multi-culturally diverse data and analyze movement signature peculiar to every tradition of dance. This makes sure that smart systems will be able to consider both universal movement properties and culturally specific aesthetic criteria.

Table 2

Table 2 Computational Methods for Expressivity Evaluation
Expressive Factor	Indicator	Computational Method	Output
Emotion	Upper-body tension, facial cues	Affective computing	Emotion probability
Musicality	Beat accuracy	Beat tracking	Rhythm score
Effort Quality	Suddenness, continuity	Laban-inspired modeling	Effort descriptors
Stage Presence	Projection, visual clarity	Silhouette analysis	Presence index

 

Feeling and expressiveness only enhance the aesthetic space. The modeling of emotional intention is supported by techniques used in affective computing, including facial-expression analysis, energy- mapping and motion-effort profiling. The movement qualities such as light, strong, sustained, and sudden, which are algorithmically categorized as inspired by Laban Movement Analysis (LMA), are found to be associated with emotional expressivity. These descriptive computations allow systems to presume the effectiveness of performance to communicate its emotional or narrative meaning. Representation of aesthetics is also carried out through the use of visual attributes Tu et al. (2020).Models assess silhouette clarity, line extension, spatial projection and overall stage presence using the analysis of video. The factors add aesthetic influence to a dancer but it should be carefully understood to prevent any prejudices connected with body shape, costume, or light. The danger of over-simplification is a serious point to be taken into account in the context of computational aesthetics. There is no single definition of aesthetic judgment with respect to culture, teaching, and choreography, so to be inclusive, algorithms should be flexible and trained on different datasets. It is not aimed at simplifying the artistic content to fixed numerical outputs but rather enhances interpretability and gives dancers meaningful context-sensitive feedback Babu et al. (2021).

5. Results and Discussion

This part contains the experimental data of the intelligent performance evaluation system used on a variety of dancers in the classics, contemporary and folk styles. Research outcomes comprise quantitative data based on AI developed technical and artistic assessment models, and then a synthesis of the data in the context of embodied practice, aesthetics theory and digital change in dance. The analysis is empirically grounded on four fundamental visualizations rhythm accuracy, artistic dimension breakdown, human-AI scoring comparison, and expressivity-smoothness relationship. Collectively, these findings indicate both the strengths and the challenges in interpretation of the use of intelligent systems in assessing performance in art.

Table 3

Table 3 Artistic Dimensions and Their Computational Indicators
Artistic Dimension	Traditional Interpretation	Computational Indicator	Example Metrics
Emotional Expression	Clarity, depth, consistency	Movement energy, facial affect	Emotion probability, valence–arousal values
Stylistic Authenticity	Adherence to genre norms	Genre-specific feature patterns	Style classification accuracy, signature vectors
Spatial Presence	Use of stage space, projection	Trajectory mapping, body expansion	Spatial spread index, reach distance
Creative Interpretation	Individuality in performance	Variation from reference patterns	Novelty score, dynamic variation

 

The rhythm accuracy increase is consistent with the literature, which implies that instant multimodal feedback (audio-visual signals and beat-corrective indicators) facilitates perceptual-motor synchronization. Dancing with the help of the system that identified off-beat deviations and gave corrective timing responses, the dancers managed to memorize rhythmic patterns more effectively. This is a point that makes you believe that intelligent systems can help enhance traditional pedagogy, but not to substitute human instruction, especially when it comes to repetitive or time-consuming training elements.

                                                                         Figure 4

The line graph represented in Figure 4, shows that there was a gradual increase in rhythm accuracy with each new training session. The upward trend demonstrates that AI-based feedback is useful to dancers as they can gradually learn to move to musical rhythms. Small variations are normal fluctuations, yet the tendency of the improvement over time proves the effectiveness of the system in enhancing the musical timing and minimizing the off-beats.

Table 4

Table 4 Key Computational Features Used in Artistic Evaluation
Feature Category	Description	Algorithms/Techniques Used	Evaluation Output
Motion Dynamics	Movement flow, continuity, phrasing	Optical flow, curvature analysis	Smoothness score, transition quality
Effort Qualities	Suddenness, tension, fluidity	Laban-inspired descriptors	Effort classification
Facial Expression	Emotional alignment	CNN-based emotion recognition	Emotion category probability
Musicality Interaction	Synchronization with rhythm	Beat tracking, DTW	Rhythm adherence score

Art quality is not restricted by technical precision but must involve expressiveness that is coordinated, style conformity, and spatial interaction. This multi-dimensional profile is represented in Figure 2 with the radar chart. The dancer showed good stylistic performance (85%), spatiality (80%), and movement quality (78%), it is possible to guess that he had a good sense of kinesthetic awareness and presence on stage. But the musicality (70%) and emotion (75%) were relatively low and this showed areas that cognitive-emotional integration might be enhanced.

                                                                         Figure 5

 

The radar chart represented in Figure 5 gives a global picture of the strengths and weaknesses of the dancer in terms of artistry. The high scores in style and spatiality underline high genre fidelity and coverage of the stage, whereas musicality is moderately lower, which indicates the opportunity to tune the rhythm. The visualization would be of great use when it comes to determining dimension-specific training requirements, so that expressive and technical areas could be specifically improved.

Table 5

Table 5 Integrated Framework for Artistic Evaluation
Evaluation Layer	Input Data	Processing Method	Artistic Output Generated
Structural Layer	Skeleton keypoints	Pose estimation, feature engineering	Technique score, alignment accuracy
Expressive Layer	Video frames, energy profiles	Emotion recognition, Laban features	Expressive clarity, emotional consistency
Contextual Layer	Audio, stage movement	Beat alignment, spatial mapping	Musicality score, spatial aesthetics
Interpretive Layer	Cross-layer fusion	Weighted scoring models	Final artistic evaluation score

The allocation supports the fact that aesthetics can be partially measurable and yet remain subjective to the artist. The capacity of the system to rate artistic attributes is based on the calculation of expressive features of motion dynamics, effort qualities, energy curves, and affective markers. Notably, according to the radar chart, it shows not only the performance results but also the training opportunities. To illustrate, a decrease in musicality implies the necessity of the exercises that entail the association of phrasing with breath-work or dynamic weighting, whereas moderate emotional clarity implies possible focus on characterizing and narrative phrasing. These results support the multi-layered evaluation model, indicating that the computational indicators of various dimensions of aesthetics are capable of representing the separate aesthetic dimensions without sacrificing the artistic uniqueness of the dancer.

                                                                         Figure 6

 

As illustrated in Figure 6, comparison indicates that there were areas of compatibility and discrepancies in human and AI assessors. The two are similar in their scoring patterns, which means that AI obtains key aesthetic cues. Minor disparities particularly in the style recognition indicate that AI can be better at recognizing patterns consistently whereas humans can emphasize more on emotional interpretation. This analogy proves the accuracy and supplementary essence of AI assessment. The critical element of the system validation is the extent of the correspondence between human experts and AI scores of the works of art. Figure 3 is a comparison of scores in emotion, style, musicality, and spatiality. The trend is very consistent with human and AI assessors with the difference falling within +-3-5%. It is important to note that AI was a little more sensitive to genre specific geometric and temporal cues than human assessors, with a higher score in style. On the other hand, human assessors rated more emotion which is in line with holistic and intuitive perception of human affect.

                                                                         Figure 7

The scatter plot provided in Figure 7, shows that there is a positive relationship between movement smoothness and the expressivity. Dancers who have smoother motions paths have a higher score in expressive clarity. Although the correlation is not ideal to show that expressive performance also consists of the emotional intention and style subtlety the overall tendency is in favor of the hypothesis that the fine motion control increases the artistic effectiveness. This correspondence represents the ability of the model to emulate the most significant aspects of art and still have a decent level of interpretability. The minor difference in human judgment and AI judgment points to the successful hybrid strategy: AI is a master at objective stylistic evaluation, whereas a human is more capable of gauging narrative and emotional complexity. Making these two views work together may result into deeper evaluation and more detailed feedback. The intelligent evaluation model is also reliable and construct valid based on these findings. The small difference in the mean score of human and AI indicates a possibility of standardising assessment criterion and still allow expert judgment. Expressive quality can be referred to as the capacity of a dancer to put life into motion. This is studied in Figure 4 by a scatter plot between smoothness and expressivity. The affirmative relationship shows that the more dancers showed the use of controlled, fluid transitions, the more expressive they were. This is in line with the Effort Theory developed by Laban where flowing movement is usually linked to an expression of openness and a clarity of emotions. Non-linear dispersion of certain data points however also points to the fact that expressivity does not solely rest on the technical smoothness. Emotions intention, story setting, and culture style play an important role in expressive output. Indicatively, abrupt shifts, tension or roundedness are used deliberately as a tool of conveying meaning in some folk and contemporary styles. Therefore, such violations highlight the necessity of culturally sensitive AI models that will be able to identify the difference between technical roughness and an intentional style.

 

6. Conclusion

This study has analyzed the dynamic nature of the intelligent performance evaluation in the training of dance where there is a convergence of the bodily artistry, the computational modeling and the digital innovation. Taking a comprehensive approach to the embodied practice, digital transformation, computational aesthetics, and AI-driven assessment of art, the study showed how technology could help dancers to perfect their technical accuracy and expressiveness. The experimental findings depicted significant changes in the accuracy of rhythm, clarity of style, and consistency of expressive features showing the possibility of intelligent feedback systems to improve the process of acquiring skills and creativity exploration. AI-based pose estimation, affective computing, immersive interfaces, and multi-modal analytics provide an excellent set of tools that can be used to understand movement quality in fine and culturally adequate terms. Although the system offers objective and data-driven information, it has value but it is supplementary to, rather than substitutive of human expertise. Human teachers introduce fundamental interpretative, emotional, and cultural knowledge which can never be replicated completely by AI. The consistency between human and AI rating also augers well the effectiveness of hybrid assessment systems which is a combination of both computational and artistic sensitivity. Simultaneously, the study also recognizes the difficulties concerning the bias of the dataset, the cultural variety, emotional interpretation, and the ethical accountability. These shortcomings will need to be tackled in order to create non-biased, culturally oriented assessment frameworks that would consider the diverse diversity of world dance cultures. Finally, it can be concluded that intelligent performance evaluation is not only a technological innovation but also a paradigm shift in the field of dance pedagogy. Through the adoption of cooperative AI-human methods, the future of dance training has the possibility of increasing access, accuracy, and creative possibilities of dancers around the world.

 

CONFLICT OF INTERESTS

None. 

 

ACKNOWLEDGMENTS

None.

 

REFERENCES

Babu, P. A., Nagaraju, V. S., and Vallabhuni, R. R. (2021). Speech Emotion Recognition System with Librosa. In Proceedings of the IEEE International Conference on Communication Systems and Network Technologies, IEEE.

Chéron, C., Laptev, I., and Schmid, C. (2015). P-CNN: Pose-based CNN Features for Action Recognition. In Proceedings of the ACM International Conference on Multimedia, ACM.

Ishii, R., et al. (2024). Let’s Dance Together! AI Dancers Can Dance to Your Favorite Music and Style. In Proceedings of the 26th International Conference on Multimodal Interaction Companion, 88–90. ACM.

Khosla, P., et al. (2021). Supervised Contrastive Learning (arXiv:2004.11362). arXiv.  

Lee, S.-E., Shibata, K., Nonaka, S., Nobuhara, S., and Nishino, K. (2022). Extrinsic Camera Calibration from a Moving Person. IEEE Robotics and Automation Letters, 7, 10344–10351. https://doi.org/10.1109/LRA.2022.3194027

Li, R., Yang, S., Ross, D. A., and Kanazawa, A. (2021). AI Choreographer: Music Conditioned 3D Dance Generation with AIST++. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), IEEE.

Lin, J. H. (2015). “Just Dance”: The Effects of Exergame Feedback and Controller use on Physical Activity and Psychological Outcomes. Games for Health Journal, 4, 183–189. https://doi.org/10.1089/g4h.2014.0100

Liu, Y. Z., Zhang, T. F., Li, Z., and Deng, L. Q. (2023). Deep Learning-Based Standardized Evaluation and Human Pose Estimation: A Novel Approach to Motion Perception. Traitement du Signal, 40, 2313–2320. https://doi.org/10.18280/ts.400621

Malleson, C., Collomosse, J., and Hilton, A. (2020). Real-Time Multi-Person Motion Capture from Multi-View Video and IMUs. International Journal of Computer Vision, 128, 1594–1611. https://doi.org/10.1007/s11263-020-01288-7

Singh, A., et al. (2021). Semi-Supervised Action Recognition with Temporal Contrastive Learning (arXiv:2102.02751). arXiv.  

Takahashi, K., et al. (2018). Human Pose as Calibration Pattern: 3D Human Pose Estimation with Multiple Unsynchronized and Uncalibrated Cameras. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops 1856–1867. IEEE.

Tsuchida, S., Fukayama, S., Hamasaki, M., and Goto, M. (2019). AIST Dance Video Database: Multi-Genre, Multi-Dancer, and Multi-Camera Database for Dance Information Processing. In Proceedings of the 20th International Society for Music Information Retrieval Conference.

Tu, H., Wang, C., and Zeng, W. (2020). VoxelPose: Towards Multi-Camera 3D Human Pose Estimation in Wild Environment (arXiv:2004.06239). arXiv.  

Ullah, I., Manzo, M., Shah, M., and Madden, M. G. (2022). Graph Convolutional Networks: Analysis, Improvements and Results. Applied Intelligence, 52, 9033–9044. https://doi.org/10.1007/s10489-022-03387-3

Yao, B., Jiang, X., Khosla, A., Lin, A. L., Guibas, L., and Fei-Fei, L. (2011). Human Action Recognition by Learning Bases of Action Attributes and Parts. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), IEEE.

Zhang, X. (2023). The Application of Artificial Intelligence Technology in Dance Teaching. In Proceedings of the 1st International Conference on Real-Time Intelligent Systems, 23–30.

Zhenyu, S. (2025). Gamified Learning Experience Based on Digital Media and the Application of Virtual Teaching Robots in Dance Teaching. Entertainment Computing, 52, Article 100785. https://doi.org/10.1016/j.entcom.2024.100785