Identification of Emotional States and Psychological Crisis Prevention for Medical Students Based on Integrated Learning and Walking Pattern

Currently, the psychological crisis prevention methods for medical students combine with artificial intelligent technology and mainly analyze students’ facial and electroencephalogram (EEG) signals. These methods have the problems of poor operability and low recognition accuracy. To address these problems, the study proposes a method for psychological state recognition and psychological crisis prevention based on ensemble learning and walking pattern analysis. Moreover, the study applies this method to the psychological crisis prevention of medical students. First, the study uses a pose estimation algorithm and a convolutional structural network to extract 3D features of walking patterns. Then, it applies these features to emotion recognition. The research selects K-nearest neighbor, decision tree algorithm, and support vector machine as the integrated weak classifier. The experimental results indicated that the recognition accuracy of the psychological state recognition method proposed in the study reached 95.63%. The response time was only 0.23 s, with a recall value of 0.96 and an root mean squared error of 0.03. In the practical application of the psychological crisis intervention, the accuracy of its early warning reached 91.48%, and the rate of improvement of the students’ psychological state after the early warning also reached 78.23%. The proposed method can help teachers gain timely insight into the psychological problems of students and carry out psychological crisis prevention.