A Quantitative Study on Students’ Motion Performance in Online Dance Course Based on Non-linear Dimensionality Reduction and Clustering

Objective: To conduct a dance evaluation model based on the online dance teaching evaluation data of physical performance and emotional expression of students. Methods: 76 college students participate in two rounds of different types of cheerleading in class and after class. According to the videos uploaded by the subjects, the score of comprehensive exercise performance, prosodic coordination, action intensity, action standard, action completeness, and the quantized values of eyes, mouth, hands, legs and psychological tension were evaluated for calculating the visual quantitative scores. The data except visual quantitative scores was clustered after non-linear dimensionality reduction and finally obtained a classification result. Results: 1) visual quantitative scores and the score of comprehensive exercise performance were correlated significantly(pom: r=0.931, P<0.01; free dance: r=0.942, P<0.01) and have no significant difference(pom: t=-1.665, P=0.098; free dance: t=0.552, P=0.581). 2) The method of non-linear dimensionality reduction and clustering can divide the samples into three distinct levels or six distinctive categories, each category has a different degree of improvement in the case of extending the same training time(1~72 h). Conclusions: The evaluation method based on non-linear dimensionality reduction and clustering has a great degree of substitutable for expert scoring. It has the advantages of fine analysis granularity, fast speed, high precision, clear training path and so on, and is applicable to various types of dances. Therefore, it is suitable for the assessment and evaluation of online dance teaching.