Classroom Behavior Analysis and Evaluation in Physical Education by Using Structure Representation

Abstract

Behavior analysis plays a critical role in physical education. This paper resorts to computer vision technology to establish a classroom behavior analysis system for physical education. First, the behavior video is collected by a Kinect camera. Then, the behavior is recognized based on the symbiotic relationship and geometric constraints between human posture and interactive objects. The human skeleton is used to describe the behavior subject and the local area boundary boxes are divided with each node in the skeleton as the center. The human posture features are used to learn a structural classification model to recognize human behavior sequence. Finally, the behavior recognition results are used to analyze physical education. The experimental results show that the proposed behavior analysis framework can accurately recognize human behavior during physical education classes.