A Study of Teacher Behavioural Pattern Recognition Techniques in Music Education and Their Impact on Teaching Effectiveness

Abstract

This paper proposes a method for recognizing teacher behavior patterns using a graph convolutional neural network. This method extracts the human skeleton point information by using a single-frame image from a teaching video as a unit. Then, a graph convolutional neural network is used as a framework to aggregate multi-frame image information and identify the teacher’s behavioral patterns. In order to verify the effectiveness of the method, the article constructs a video dataset collected from public music teaching videos on Catechism.com, which contains six categories of teacher behavioral patterns and conducts comparative experiments. The experimental results show that the teacher behavior pattern recognition technique based on a graph convolutional neural network can effectively exclude the interference of irrelevant information within the classroom scene and make full use of the information generated between the skeleton points in the multi-frame image to accurately identify the typical teaching behaviors of the teacher, with an accuracy of more than 0.96 and stronger robustness. This study can respond to teachers’ teaching status in a timely and effective manner, which can help teachers optimize their teaching behaviors and improve their teaching effectiveness in music classrooms.