Continuous Wavelet Transform Analysis for the Classification of Surface Electromyography Signals

A number of Digital Signal Processing (DSP) techniques are being applied to Surface Electromyography (SEMG) for classification using signal feature extraction. This research is aimed at using Continuous Wavelet Transform (CWT) analysis of SEMG signals to develop and adopt a sound methodology for classification of the signals at different force levels. Traditional analysis methods such as Fast Fourier Transform (FFT) could not be used alone because muscle diagnosis requires time-based information. Therefore CWT was selected for this research as it includes time-based information as well as scales that can be converted into frequencies, making muscle diagnosis easier. CWT produces a scalogram plot along with its corresponding time-based frequency spectrum plot. Using both of these plots, extracted features of the dominant frequencies and the related scales can be used to train and validate a signal classifier based on an Artificial Neural Network (ANN). SEMG signals were obtained for a 10 second period sampled at 2048 Hz and digitally filtered using a Butterworth bandpass filter (5 to 500 Hz, 4th order). Signals were collected from the vastus medialis muscle of both legs of 45 healthy male subjects at 25%, 50% and 75% of their Maximum Voluntary Isometric Contraction (MVIC) force of the quadriceps. The extracted features selected for the two second period of the signal were the mean and median frequencies of the average power spectrum, and the RMS values at scale 8, 16, 32, 64 and 128 of the scalogram. The signals were analysed using CWT in LabVIEW with its Signal Processing Toolset. In this paper, the results are presented to show that the selected extracted features are suitable for the classification of SEMG signals at different force levels.