Classification of EMG Finger Data Acquired With Myo Armband

Abstract

Muscles are one of the basic building blocks of our body, which allows us to perform various movements. As a result of the contraction and relaxation of the muscles, myoelectric signals are formed and movement is provided. EMG signal is obtained by measuring these signals with electrodes. By processing EMG signals, human movements can be imitated and used in many different areas.The human hand can perform many combinations of hand gestures thanks to the different mobility of the fingers. For this reason, it can be easier for the prosthetic hands to perform different hand gestures by moving the fingers independently. In this context, the aim of the research is to propose a model by processing EMG signals so that finger gestures can move independently. EMG signals were acquired using myo armbands. Data set was created with 5 finger gestures and resting hand gestures. The data set was filtered, the part where the gesture was performed in the preprocessing was determined and the windowing process was applied. The classification process was performed by eliminating the features of the EMG signals. In the 100 ms 50% overlapping window, 95.8% classification success was achieved by using the SKNN method with the EWL feature. When the experimental results were examined, it was observed that a successful model was created.In this study, a model for prosthetic arm control was proposed by processing finger data. This model is feasible for prosthetic arms and it is estimated that the functionality of the prosthetic arm will be increased by processing finger data. © 2020 IEEE.