sEMG Signal Enhancement using Cubical Denoising for Wrist movement Classification

This paper presents the developments made in the research for classifying different wrist movements using surface Electromyogram (sEMG) signals. The strategy discussed within uses the concepts of pattern recognition to classify different classes as wrist movements. In order to minimize the effect of noise involved with sEMG during recording, wavelet denoising is implemented using Daubechies mother wavelet. It employs a cubical function for soft thresholding which has provided with finest results as compared to the previous researches. Spectral features, Reflection coefficients along with Wilson's amplitude and other features were extracted and provided to the linear classifier. The results calculated from these experiments indicate better recognition performance using the given features when denoising is applied. The maximum classification accuracy obtained for the identification of four wrist movements was 98.5% which is quite significant as compared to the previous researches.