AI-based Multimodal Stress Detection and Physiological Tracking Application

This study presents StressEase, an AI-powered platform for real-time stress detection using three data modalities: facial expressions, speech sentiment, and smartwatch physiological signals. The system aims to offer reliable, non-invasive stress monitoring through machine learning and deep learning techniques. A Convolutional Neural Network (CNN) is employed for facial expression analysis, achieving 90% accuracy. For speech sentiment analysis, an SVM classifier with TF-IDF features is used, achieving 96% accuracy. Smartwatch data, including heart rate and sleep patterns retrieved using the Google Fit API, is processed by an XGBoost model with 92% accuracy. These outputs are fused through an ensemble strategy to determine the overall stress level. The system is deployed as both a web and mobile application, developed using React for the front end, Flask for the backend, and SQLite for the database, offering real-time insights and personalized relaxation recommendations. The ensemble model's combined predictions demonstrate the system's high reliability and potential for real-world applications in health, workplace wellness, and mental health monitoring.