EMG Signal Classification Using Radial Basis Function Neural Network

Abstract

Classification of electromyography (EMG) signals of human arm using Radial Basis Function Neural Network (RBFNN) is presented. Using an 8-channel Myo armband, EMG signals are collected from the arm muscles. Several time domain features are extracted from the collected EMG signals; then an RBFNN is trained. In the training process, the patient moves his\her hand according to a predefined position to obtain training patterns. After the training process is completed, the trained RBFNN can recognize the patient’s intended gesture from the hand’s EMG signals and consequently the patient can control a prosthetic hand’s movements. One RBF network is trained for each extracted feature, and excellent classification results are achieved. Next, different structures of RBFNN are implemented to obtain a simpler classifier. A MATLAB program is written to train networks and record the results. The experimental results show that RBFNN is an excellent classifier with RMS error less than or equal to 10-15 for implementing a myoelectric prosthetic hand. This work is carried out by Intelligent Prosthetics Research Group (IPRG) at College of Engineering, University of Kerbala in 2018.