Wavelet Transform to Recognize Muscular: Force Relationship Using sEMG Signals

Abstract

Being man–machine interaction, the use of surface electromyogram (sEMG) is increasing day by day. Generally, sEMG is a complex signal and is influenced by several external factors/artifacts. As removing these artifacts is not easy, feature extraction to obtain useful information hidden inside the signal becomes a different process. This paper presents methods of analyzing sEMG signals using discrete wavelet transform for extracting accurate patterns of the sEMG signals. The results obtained suggest having a good compromise between the percentage root mean square differences, root mean square difference value for the denoising and quality of reconstruction of the sEMG signal. Further a one way separated factorial analysis was performed to find out the effectiveness of analyzed sEMG signal for discrimination among different classes of groups Various possible types of wavelets with high level parameters were tested for denoising and results show that the best mother wavelets for tolerance of noise are fifth order of symmlets and bior6.8 whereas for reconstruction, wavelet functions bior5.5 and sym3 were the best.