An Early Warning Method of College Students’ Psychological Crisis Based on Emotion Recognition

Abstract

In order to realize rapid and accurate early warning of college students’ psychological crisis, an improved emotion recognition method based on the YOLOv8 model is proposed. Firstly, a depth-separable deep convolutional module is introduced based on YOLOv8 algorithm to improve the detection effect of facial expression details. Secondly, a new detection frame regression loss is introduced to improve the accuracy of college students’ emotion recognition. Finally, the effectiveness of the proposed method is verified on a standard emotion recognition public dataset. The experimental results show that compared with several variants of YOLOv8, our model is significantly superior to other comparison methods and the overall accuracy of emotion recognition reaches 0.784. Our method has better stability and sustainability. The accurate basis can be provided to college students’ psychological early warning.