Wavelet analysis-based evaluation of electromyogram signal using human machine cooperation

Abstract

The human body is a combination of interacting systems that can be analysed using engineering principles. It is well known that surface electromyogram signals easily acquired from surface of skin of the body using non-invasive electrodes are composed with variety of noises. Hence methods to remove noise become most significant for surface electromyogram (sEMG) signal before performing processing and analysis. In this study, wavelet analysis has been used to analyse quality of effectiveness of surface electromyogram signal. The surface electromyogram signals were estimated with the following steps: first, the obtained signal was decomposed using wavelet transform; then, decomposed coefficients were analysed by threshold methods. Daubechies wavelets (db2-db14) family for efficiently removing noise from the recorded surface electromyogram signals has been used. However, the most essential wavelet for surface electromyogram denoising is chosen by calculating the root mean square value and signal power values from different voluntary contraction motions. The combined results of root mean square value and signal power shows that wavelet db4 performs denoising best among the wavelets. Furthermore, the statistical technique of analysis of variance (ANOVA) for experimental and best wavelet coefficient was analysed to investigate the effect of muscle-force relationship for ensuring class separability.