VISUALSTRESS: AN AI-BASED APPROACH TO HUMAN STRESS DETECTION THROUGH IMAGE ANALYSIS

Abstract

Stress has become a pervasive factor impacting human health, productivity, and overall well-being. Traditional stress assessment methods, such as self-reports and physiological sensors, are often intrusive, subjective, or impractical for continuous monitoring. Recent advancements in Artificial Intelligence (AI) and computer vision have opened new opportunities for non-contact, automated stress detection through image analysis. This study proposes VISUALSTRESS, an AI-based framework that leverages facial expressions, micro-expressions, remote photoplethysmography (rPPG), and thermal imaging cues to identify stress states in real time. The system integrates deep learning models for facial feature extraction, temporal pattern recognition, and physiological signal estimation from videos, enabling accurate detection of stress without requiring physical sensors. Experimental evaluations on benchmark datasets and controlled stress-induction protocols demonstrate the framework’s ability to distinguish stress levels with high reliability, outperforming traditional handcrafted feature-based approaches. By providing a non-invasive, scalable, and privacy-aware solution, VISUALSTRESS has potential applications in healthcare, workplace wellness, driver monitoring, and human-computer interaction. Furthermore, the approach addresses challenges of robustness, individual variability, and real-world deployment, paving the way for next-generation intelligent stress detection systems.