Towards Analyzing Psychomotor Group Activity for Collaborative Teaching Using Neural Networks

Collaborative teaching in psychomotor skills requires the ability to detect and analyze the movements and postures of multiple participants simultaneously. In this paper we explore the feasibility of analyzing group activity in psychomotor contexts using neural networks. We apply our approach to kihon kumite, a training exercise in karate where two participants perform predetermined techniques. The overall activity is measured as the sumof individual performances, with the support of computer vision techniques to label individuals and classify postures. Furthermore, interactions between participants are also analyzed, which allows for a more comprehensive understanding of the group activity. Our approach shows promising results for improving collaborative teaching in psychomotor skills and can be extended to other physical activities.