Automatic Recognition of Teachers' Nonverbal Behavior Based on Dilated Convolution

Abstract

Recognition of teachers' nonverbal behavior is an important tool for assessing teachers' teaching behavior. Within the continuous progress of artificial intelligence technology, automatic recognition of nonverbal behavior has been more and more popular. Therefore, we take the teaching videos in the actual classroom as the data source, acquires and labels the teachers' nonverbal behavior dataset, extracts the teacher skeleton points information in the images using the target detection algorithm and the human pose estimation algorithm, and adopts a residual block model with dilated convolution for automatic recognition of teachers' nonverbal behavior. It is proved that the proposed method can effectively improve the accuracy of teachers' nonverbal behavior recognition. The research in this paper is to help solve the problem of automatic recognition of teachers' nonverbal behavior and assist teachers to optimize teaching strategies and improve teaching efficiency.