Learning behavior recognition and teaching effectiveness analysis in smart classrooms for international Chinese education

This paper briefly introduces the learning behavior recognition algorithm for smart classrooms. This algorithm first uses You Only Look Once version 5 (YOLOv5) to identify and locate students in classroom surveillance images. It then utilizes OpenPose to extract the student action skeleton within the localization box. Finally, a convolutional neural network (CNN) is used to classify the learning behavior corresponding to each extracted skeleton. A case analysis was conducted to evaluate the algorithm. First, the proposed algorithm was compared with the traditional CNN, YOLOv5, and Faster region-based convolutional neural network (R-CNN) algorithms to verify its performance in recognizing learning behaviors. Then, its effectiveness in Chinese teaching was verified using 60 non-native Chinese speakers. The results indicated that the learning behavior recognition algorithm had better recognition performance compared to the traditional CNN, YOLOv5, and Faster R-CNN algorithms. When it was applied to the teaching of the international Chinese smart classroom, there was a significant improvement in students’ Chinese scores compared to the traditional teaching mode.