Low cost solutions based on wearable devices to measure fatigue in workers

Every year, many workplace accidents occur due to workers’ accumulated fatigue, especially in jobs that require a high level of physical exertion, resulting in stressful environments. This research fits into this contest, with the goal of monitoring physical fatigue: laying the groundwork for developing a device able to determine when a person is too fatigued to continue working safely. To create a controlled environment that induces physical fatigue, each of the 30 volunteers underwent an incremental exercise protocol designed to induce cardio-respiratory fatigue. The protocol was performed on a Garmin bicycle (Tacx Neo Bike Plus) and monitored via a Polar Chest strap (Polar H10). At each stage of the exercise, participants were asked to complete the Rate of Perceived Exertion (RPE) scale to obtain a subjective evaluation of fatigue. Regarding data analysis, an adaptive filter algorithm was applied to RR intervals to mitigate artifacts using a threshold range of 280-1200 ms (corresponding to 210-50 bpm). Artifacts were identified and replaced using an adaptive estimation of the mean and standard deviation, preserving physiological variability while minimizing nonphysiological distortions. Subsequently, Detrended Fluctuation Analysis (DFA), a nonlinear method for analyzing fractal signals such as heart rate variability (HRV), was applied. Since HRV exhibits rapid fluctuations due to autonomic nervous system activity and slow oscillations influenced by circadian rhythms and other physiological adaptations, traditional linear analysis methods are limited in distinguishing these components. In contrast, DFA applies detrending at different scales, allowing HRV fluctuations to be more effectively analyzed. By applying DFA to measure short-term correlation persistence (scales from 4 to 16 beats), the α1 value ranged approximately from 1 to 1.4 for all subjects at rest. As the Rate of Perceived Exertion (RPE) increased, α1 gradually decreased, reaching a minimum of around 0.5 at maximal fatigue, followed by a rapid increase during the recovery phase. The results show an inverse relationship between perceived fatigue (RPE) and α1: as subjective fatigue perception increases, α1 decreases. The findings suggest that DFA applied to HRV is a promising approach to detect acute physical fatigue. This research represents a first step in the algorithm development of low-cost wearable devices aimed at monitoring and predicting worker fatigue, with potential applications for workplace safety. Future studies could extend this methodology to other contexts, focusing particularly on fatigue in different industrial sectors.