Detection of EMG Myopathy Signal Using Wavelet Transform and Neural Network Techniques
Abstract
This paper recognizes signals from two sources, where one is a normal person and the other is a myopathy patient. The signals under experimentation were acquired from the branchial biceps (BB) muscles using a needle electrode. This paper focuses on the technique of Wavelet Transformation (WT) to extract the features from an EMG (Electromyogram) signal. It includes decomposing the EMG signal into different levels to generate the coefficients. Self-Organising feature map (SOFM) and Linear Vector Quantisation (LVQ) neural networks were constructed from the extracted coefficients and both of these were trained with supervision. These two networks proved to be a powerful tool for diagnostic purposes, clearly separating normal EMG from a myopathic one.
Keywords
Electromyogram, Myopathy, Wavelet Transformation, Self-Organising Feature Map, Learning Vector Quantisation.