Research on activity recognition of students in physical education based on IDT method

With the advancement of new curriculum reform, educators put more attentions on the quality of physical education (PE). However, the traditional way to evaluate the PE class is not efficient and often time-wasting. In this paper, we attempt to realize the automatic evaluation of PE class by activity recognition. Recently, activity recognition methods are widely applied in various fields, e.g., surveillance, content-based video indexing. This paper utilizes an activity recognition method to evaluate the performance of students in PE from their activities recorded in videos in order to realize intelligent evaluation of the physical education. In this process, the improved dense trajectory (IDT), fisher vector (FV), principal component analysis (PCA) and support vector machine (SVM) are in use. The experiments demonstrate that no matter what feature descriptor is used, the classification accuracy is adequate, so that the activity recognition is effective in PE and the quality of PE can be evaluated automatically.