Aiming at the problems that movement evaluation in Chinese dance teaching relies on empirical observation, feedback lags and is difficult to quantify, this paper designs a performance movement optimization and teaching feedback system based on motion capture model. The system takes multi-view video as input, combines two-dimensional pose estimation, three-dimensional skeleton recovery, spatio-temporal feature modeling and deviation semantic mapping, and realizes the integrated processing of action recognition, quality assessment and correction suggestion generation. The experimental results show that the recognition accuracy of the proposed method on the self-built dataset reaches 94.82%, and Macro-F1 reaches 94.17%. After the system assisted training, the joint Angle error, trajectory deviation, rhythm deviation and center of gravity stability deviation are significantly decreased. This method can improve the refinement level of Chinese dance performance movement recognition, deviation diagnosis and classroom feedback, and provide an implemensible technical path for Chinese dance digital teaching.
