Remote assessment of physiological parameters by non-contact methods to detect mental stress

Psychological stress recognition through facial analysis has gained increasing attention due to stress-induced alterations in facial pulse patterns and micro-expressions. In this paper, we propose a dual-module framework integrating pulse signals and micro-expression features for stress detection. The pulse detection module recognizes the presence or absence of pressure based on the pulse in the facial video, and the image detection module predicts the pressure based on the facial expression features of the face. The model further combines the predicted values of facial pulse and facial image at the decision level for final decision making. Evaluated on 312 videos from 52 subjects (67% stressed cases), our method achieves 80.13% accuracy and 72.46% F1-score, outperforming single-modality pulse-based approaches. This multimodal solution demonstrates potential for real-time stress monitoring applications.