Surface Electromyography based Detection of Neck and Shoulder Muscle Fatigue using Maximal Overlap DWT and Entropy Features

Abstract

Extended use of computers, laptops, and mobile devices is a major contributor to musculoskeletal disorders, often leading to neck and shoulder fatigue. Muscle fatigue is characterized by reduced force-generating capacity and altered neuromuscular activity, typically assessed using surface electromyography (sEMG). However, the nonlinear and nonstationary nature of sEMG signals makes its analysis challenging. This study investigates detection of muscle fatigue in neck and shoulder muscles using the Maximal Overlap Discrete Wavelet Transform (MODWT). The sEMG signals were acquired from the Sternocleidomastoid (SCM), Splenius Capitis (SPC), and Trapezius (TPZ) muscles of fifty volunteers. Following preprocessing, Shannon Entropy (SE) and Permutation Entropy (PE) were extracted from sevenlevel MODWT sub-bands to capture signal irregularity and temporal complexity. Entropy features are found to be distinct across both conditions in all muscles (p<0.05). Fatigue is particularly associated with increased irregularity and altered temporal dynamics in the mid-frequency bands (D4-D6), where entropy measures serve as robust biomarkers. Classification was performed using Support Vector Machine (SVM) and Convolutional Neural Network (CNN), with CNN exhibiting superior discriminative capability and achieving an accuracy of 89.00%. The results demonstrate that MODWTbased entropy features provide reliable biomarkers for monitoring posture-induced muscle fatigue.