Characterizing Human Limb Movements Using An In-House Multi-Channel Non-Invasive Surface-EMG System

Abstract

Electromyography (EMG) signals are obtained from muscle cell activity. The recording and analysis of EMG signals has several applications. The EMG is of diagnostic importance for treating patients suffering from neurological and neuromuscular disorders. Conventional methods involve placement of invasive electrodes within the muscles to record EMG signals. The goal is to showcase the usage of surface based EMG signals to characterize all possible human limb movements. An in-house non-invasive EMG signal acquisition system that offers characterization of human limb actions is a suitable candidate for motor impairment studies and easily extendable to design bionics control specifically for neuromuscular disorder patients. An in-house 8-channel surface-EMG signal acquisition system was designed, fabricated, and employed for characterizing specific movements of upper and lower limb. The non-invasive acquisition system captures the compound electromuscular activity generated from the group of muscles. The EMG acquisition system was designed as a modular structure where the front end analog circuit designs were replicated for all 8 channels, and were designed to function independently. Support vector machine (SVM) as classifier models were developed offline to successfully characterize different human limb actions. The in house built 8 channel acquisition system with ML classifier models were utilized to successfully characterize movements at various joints of the upper and lower limb including fingers, wrist, elbow, shoulder, knee, and ankle individually.