Performance Assessment of the Optimum Feature Extraction for Upper-limb Stroke Rehabilitation using Angular Separation Method

Abstract

Most of the human everyday activities will require the use of their upper-limb muscles. The pattern of upper-limb muscle movement can be used to estimate upper-limb motions. Fundamental arm movement which is part of upper-limb muscle rehabilitation activity has been studied in order to investigate the time domain features, frequency domain, and time-frequency domain from the surface electromyogram (sEMG) signal of the upper-limb muscle. The relationship of electromyogram (EMG) signal and the rehabilitation exercise of related upper limb muscles movements are analyzed in this study. Then the features from the three domains were compared using Angular Separation Method to determine optimal feature. The result shows that MinWT has the best value of similarity which is 0.98, followed by a MeanWT feature which resulted in 0.91 of similarity. These results of EMG signal feature extraction can be used later in the study of human upper-limb muscle especially for analyzing EMG signal from patient undergone a rehabilitation treatment.