An action analysis algorithm for teachers based on human pose estimation

Teachers serve as transmitters of cultural knowledge and play a crucial role in ongoing development and continuity of human society. Analyzing their behavior can assist educators in identifying their strengths and weaknesses within the classroom, thereby enhancing their teaching skills and methods. With the development of artificial intelligence, it is possible to automatically analyze teachers' behavior in the classroom by offering timely feedback and suggestions to improve teaching strategies. In this paper, we propose an action classification method based on human skeletal posture to analyze teacher-behavior in the classroom. First, by using the HRNet pose estimation algorithm to extract teacher skeleton information as features, the algorithm could remove redundant information in images and accurately reflect the teacher's posture. Second, by considering the relative positions of skeletal points, we establish a set of quantifiable indicators to analyze the interrelationships among these points and then obtain the teacher's local posture. Finally, relying on the local posture obtained based on the indicators, we encode and classify common behaviors of teachers in the classroom. The approach's effectiveness is evidenced by experimental results, which enables dynamic management and assessment of teacher behavior in educational environments.