DensePhysioNet: AI-Driven Stress Detection Model using Physiological Dynamics

Stress is a complex mind-body burden affecting health, life quality, and productivity, and underpins many physical and mental disorders. This paper describes an AI-based multimodal framework for stress detection from synchronized ECG, EMG, and BVP signals of the WESAD dataset that supports modelling cardiovascular, muscular, and vascular stress dynamics. After processing more than 60 million samples into 37,498 multimodal windows, the DensePhysioNet model achieves 98.0\% accuracy, AUC 0.995, and an error rate of 0.0199, establishing its strong generalizability in effective multimodal learning. Stress predictions emerge 30 seconds in advance with interpretable physiological indicators and clinical-style alerts. Applications that could be developed based on this work include workplace mental health monitoring, remote patient assessment, high-risk occupational stress management, and next-generation wearables. The framework further supports Sustainable Development Goals (SDGs) in Good Health and Well-Being (SDG 3) by early detection and prevention of stress and reduction of long-term mental health burdens, and decent work and economic growth (SDG 8) through healthier and productive environments and reduction of burnout.