Real-Time Monitoring and Intervention Framework for College Students' Mental Health Based on Multimodal Interaction

Abstract

This study addresses the mental health issues of college students by proposing a multimodal interaction framework. Through multimodal data collection and integration, deep learning algorithm development, context-aware technology application, and rigorous experimental design, real-time monitoring and personalized intervention for mental health status were achieved. The results show that the multimodal data fusion framework based on the CNN-LSTM model achieves a monitoring accuracy of up to 93.8%, significantly outperforming traditional unimodal methods. The experimental group demonstrated a significant improvement in mental health, with an average reduction of 12% in their SCL-90 scores, notably better than the control group. Additionally, the system received a user satisfaction score of 8.7/10, with an average usage frequency of 3.8 times per day, validating its user acceptance and stickiness. This study provides an innovative model for mental health services in universities, with significant theoretical and practical implications.