Perceived Fatigue Progression Tracking During Manual Handling Tasks Using SEMG Recordings

Physical fatigue significantly contributes to work-related musculoskeletal disorders, highlighting the need to understand its effects during manual handling tasks for effective prevention strategies. This study examines the correlation between changes in the myoelectric manifestation of fatigue (MMF) indicators and participants’ perceived exertion during prolonged manual handling tasks using surface electromyography (sEMG) sensors. Given that the task involves various activities with different muscle engagements and ranges of motion, joint angles were obtained using inertial measurement units and used to segment the sEMG recordings based on activity and the joints’ range of motion. Linear and complexity-based MMF indicators were then extracted from these segments, and their correlation with perceived exertion was evaluated. Linear indicators, such as activation level and median frequency, showed significant correlations with perceived fatigue (p < 0.05) in the lower leg muscles, including Lateral Gastrocnemius and Tibialis Anterior, with inconsistent results in other muscles. In contrast, complexity-based indicators, including mobility, fuzzy entropy, and Dimitrov’s index, demonstrated significant correlations (p < 0.05) in eight out of ten studied muscles across all activities), revealing reduced signal variability, increased irregularity and shifts in spectral properties as fatigue progressed. Finally, a deep learning model was developed, achieving 69% accuracy for a five-stage fatigue classification using MMF indicators. Our study showed the greater potential of complexity-based MMF over linear ones for perceived fatigue monitoring during manual handling tasks and the suitability of MMF indicators for classifying perceived fatigue stages using machine learning algorithms, potentially reducing the occurrence of severe fatigue in workplace settings.