Quantification of sEMG Signals for Automated Muscle Fatigue Detection Using Nonlinear SVM

Fatigue is a multidimensional and subjective concept and is a complex phenomenon including various causes,mechanisms and forms of manifestation. Thus, it is crucial to delineate the different levels and to quantify selfperceivedfatigue. The aim of this study was to introduce a method for automatic quantification and detection ofmuscle fatigue using surface EMG signals. Thus, sEMG signals from right sternocleidomastoid muscle of 9 healthyfemale subjects were recorded during neck flexion endurance test in Quaem hospital. Then six features in time,frequency and time- scale domains were extracted from signals. After dimensionality estimation and reduction, theSVM classifier was applied to the resulted feature vector. Then, the performance of linear SVM and nonlinear SVMwith RBF kernel and the effect of value in RBF kernel, on the accuracy of classification were evaluated. The resultsshow that the best accuracy is achieved using RBF kernel SVM with equal to 0.5 (91.16%) and also the selectedfeatures using LLE criterion, were RMS, ZC and AIF. These results suggest that the selected features contained someinformation that could be used by nonlinear SVM with RBF kernel to best discriminate between fatigue andnonfatigue stages.

Keywords:

Surface Electromyography (sEMG) , SternoCleidoMastoid muscle (SCM) , muscle fatigue , classification , Radial Basis Function (RBF) kernel , Support Vector Machines (SVM)