Research on Data-Driven Teaching Evaluation Methods for Dance Education Intelligent Platforms

Abstract

In this paper, a Springboot-based online learning service visualization and analysis system is first developed to show the overall situation of the dance education intelligent platform through visual analysis. A camera-aware Z-regression regression module is designed to realize gesture-based dance movement recognition. Unsupervised motion recognition based on S-SOFM neural network is proposed to complete dance movement evaluation using OE-DTW algorithm. The advantages of this system are verified through comparative tests and system application analysis. At the level of dance movement recognition, Z-regression performs optimally in all scenarios, and in the JHMDB dataset, the recognition rates of upper body, lower body and whole body reach 69.4%, 63.9% and 70.1%, respectively, which are significantly higher than that of RGB and Flow. Combined with the ratings of experts in the field of dance, the average Spearman correlation coefficient, ρ, of the system in this paper reaches 97.1%, which is able to effectively evaluate and analyze the dance movements.