A method for dance therapy movement data segmentation and emotion recognition based on movement rhythm and dynamic time warping

Iranian Journal of Science and Technology, Transactions of Electrical Engineering · Springer Nature · 2026-11-19 · DOI: 10.1007/s44163-026-02380-z

Аннотация

This paper constructs a comprehensive framework for dance therapy movement segmentation and emotion recognition. This framework primarily addresses the limitations of traditional dance movement emotion assessment. We first use standardized preprocessing and global noise reduction on the raw data to construct multidimensional motion representations, including velocity, acceleration, and joint velocity. Based on this, we build an adaptive dual-threshold rhythm segmentation algorithm. This algorithm can periodically statistically analyze the whole-body motion energy and automatically detect emotional units, obtaining their natural boundaries. This method overcomes the integrity degradation inherent in traditional fixed-time-window methods. Furthermore, we incorporate a time-planning classifier, which can handle different feature dimensions and assign discriminative weights. These weight variations allow for flexible weight allocation for different motion units and dynamic attention to emotion classification. We constructed a dataset containing five emotion categories: joy, sadness, anger, fear, and calmness. The dataset shows an overall prediction accuracy of 82% and an F1 score of 80%. Compared to fixed-window methods, our model demonstrates a significant performance improvement. Moreover, the model can not only generate discrete labels but also construct continuous emotion sequences. This result provides dance therapists with a visualized process. Analytical tools advance dance movement therapy towards a more precise, intelligent, and data-driven approach.

Читать у издателя →

Похожие статьи