The traditional dance teaching method has subjective evaluation, different evaluation standards, etc., which cannot provide learners with accurate movement skill levels, so a movement skill level analysis model based on computer vision and deep learning algorithms is proposed to improve the professional degree of dance teaching. Using a multi-source information fusion method, a color depth camera and an inertial measurement unit are used to collect human movements in real time and analyze them based on a spatio-temporal convolutional neural network combined with a multi-scale attention mechanism for accurate posture estimation and movement evaluation. The results of this study show that the system has achieved good results in terms of correctness of movement, reaction speed, and experience, and the results of an eight-week controlled trial show that the system can effectively help dancers improve their skill level, fluency, rhythm, and expressiveness. According to the feedback information from users, the real-time feedback function and personalized learning program design of the system are well received by users.
