Variational mode decomposition based differentiation of fatigue conditions in muscles using surface electromyography signals

Surface electromyography (sEMG) signals are stochastic, multicomponent and non-stationary, and therefore their interpretation is challenging. In this study, an attempt has been made to develop an automated muscle fatigue detection system using variational mode decomposition (VMD) features of sEMG signals and random forest classifier. The sEMG signals are acquired from 103 healthy volunteers during isometric (45 subjects) and dynamic (58 subjects) muscle fatiguing contractions and preprocessed. The band-limited intrinsic mode functions (BLIMFs) are extracted from non-fatigue and fatigue segments of the signals using the VMD algorithm. Hjorth features, such as activity, mobility and complexity are extracted from each BLIMF and are given to the random forest classifier. The performance of these features is evaluated using leave-one-subject-out cross-validation. The results show that the complexity feature performs better than others and it has resulted in an accuracy of 83% in dynamic contractions and 80% in isometric contractions. The performance is increased by about 8% in a dynamic condition when the most significant complexity features (p < 0.001) are used and by about 12% for isometric when the authors use all significant features. Therefore, the proposed approach could be used to detect fatigue conditions in various neuromuscular activities and real-time monitoring in the workplace.