Detection method of students' English classroom learning behaviour: multi-channel feature fusion

Aiming to improve the problems of large root mean square error and low F1-value in existing methods, a student English classroom learning behaviour detection method based on multi-channel feature fusion is proposed. Firstly, collect classroom data such as students' oral pronunciation characteristics, note taking texts, eye tracking data, and facial expression recognition. Secondly, extract student English classroom learning behaviours, including oral pronunciation features, text features, and visual attention features; then, input the above features into a recurrent neural network to achieve feature fusion. Finally, establish a machine learning model, use semi supervised learning for model training, and use the trained model to detect student English classroom learning behaviour. The experimental results show that the average RMSE of the proposed method is 0.30, and the F1-value is higher, fully indicating that its detection effect is better.