Wu, Aizhe
Guo, Jing
Wang, Kai
Yang, Shuai
Ke, Yongzhen
Article History
Received: 23 December 2025
Revised: 19 January 2026
Accepted: 8 February 2026
First Online: 19 February 2026
Declarations
:
: The research uses publicly available datasets (TikTokDance5K, TikTokDataset) and a self-built dataset (FC-100) collected from legal and public online sources. No human subjects were directly involved in the study, and no personal sensitive information (e.g., private portraits, personal identifiers) was included in the datasets. Therefore, ethical approval and consent to participate are not applicable.
: The datasets used in this study do not contain any identifiable personal information, and all content is derived from public resources. Consent for publication is not required.
: All authors declare no conflicts of interest. No financial or non-financial competing interests exist that could influence the design of the study, interpretation of results, or publication of the manuscript.
: All materials (including dataset annotations, preprocessing scripts, and model checkpoints) necessary to reproduce the study results are either publicly accessible via the code repository or available from the corresponding author upon reasonable request, with no restrictions.
: All training/testing code, model configurations, pose extraction scripts, and evaluation metrics code are publicly accessible at: The repository includes detailed instructions for reproducing the experiments, ensuring full reproducibility.
: The authors declare no competing interests.