Evaluating the Potential of Wearable Technology in Early Stress Detection: A Multimodal Approach

Abstract Stress can adversely impact health, leading to issues like high blood pressure, heart diseases, and a compromised immune system. Monitoring stress with wearable devices is crucial for timely intervention and management. This study examines the efficacy of wearable devices in early stress detection using binary and five-class classification models. Significant correlations between stress levels and physiological signals, including Electrocardiogram (ECG), Electrodermal Activity (EDA), and Respiration (RESP), were found, validating these signals as reliable stress biomarkers. Utilizing the WESAD dataset, we applied ensemble methods, Majority Voting (MV) and Weighted Averaging (WA), achieving maximum accuracies of 99.96% for binary classification and 99.59% for five-class classification. Ten classifiers were evaluated, with hyperparameter optimization and 3 to 10 fold cross-validation applied. Time and frequency domain features were analyzed separately. We reviewed commercially available wearables supporting these modalities and provided recommendations for optimal configurations in practical applications. Our findings demonstrate the potential of multimodal wearable devices for early detection and continuous monitoring of psychological stress, suggesting significant implications for future research and the development of improved stress detection systems.