A Method of Psychological Stress Recognition Based on Pulse Rate Variability Signals and Deep Learning Technology

Chronic psychological stress can not only lead to mental disorders but also trigger a range of physical illnesses, posing serious threats to both mental and physical health. Timely and effective psychological intervention is essential to reduce the risk of progression to anxiety or depression. Therefore, accurate identification of psychological stress is of great importance. This study proposes a novel method for psychological stress assessment based on photoplethysmography (PPG) signals. Specifically, pulse rate variability (PRV) signals are first extracted from raw PPG data to capture key physiological features that reflect autonomic nervous system activity. A deep learning framework, termed MSRC-BiLSTM, which integrates Multi-Scale Residual Convolution with Bidirectional Long Short-Term Memory networks, is then constructed to efficiently learn the spatiotemporal features of PRV signals. Finally, a fully connected neural network is employed to classify psychological stress states. To evaluate the effectiveness of the proposed method, experiments were conducted on the public WESAD dataset. The results show that the proposed method achieves superior performance compared to existing approaches, with an accuracy of 93.60% and an F1-score of 90.14%, demonstrating its significant advantages in stress recognition tasks. This method not only provides valuable support for clinicians in formulating personalized mental health interventions but also serves as an objective tool for evaluating the effectiveness of psychological stress treatments, showing promising application potential.