Abstract
Workplace fatigue caused by stress is a significant concern in physically demanding environments such as construction sites, where long hours and strenuous labor increase the risk of injuries and health issues. To address this challenge, we developed a wearable, user-friendly stress sensor capable of monitoring cortisol levels in human sweat-a key biomarker of physiological stress. The sensor incorporates a functionalized aptamer-based assay to enable real-time, on-site tracking of fatigue symptoms. Impedance-based measurements, specifically changes in charge transfer resistance, were recorded from 10 healthy individuals during a physical stress-inducing hammering task. The sensor showed an increase in charge transfer resistance following the onset of the hammering activity, indicating the physical stress experienced by the participants. Furthermore, cortisol sensor data, combined with other physiological metrics like heart rate variability, were incorporated into a machine learning framework to classify participants' fatigue levels. Of the various models evaluated, the support vector machine achieved the highest classification accuracy of 0.764, effectively distinguishing between low, medium, and high fatigue levels based on quantile-based distributions of signal changes. Hence, this work advances the development of a wearable platform for continuous stress and fatigue monitoring, supporting occupational health management in high-risk labor environments.