WAVELET ENERGIES AS A FEATURE AND THEIR IMPACT ON CLASSIFYING MOVEMENTS BASED ON sEMG

摘要

Surface Electromyography signal processing and classification is an issue that concerns a large number of research groups, demanding more accurate, simple and sophisticated feature extraction schemes in order to accomplish better performance in different applications, with a solid subject being the control of prosthetics since decades ago with early signs of satisfying accuracy. In this research, we investigate the effect of efficient feature extraction on the wavelet domain using the discrete Wavelet transformation (DWT), on NINAPPRO, a database of 27 subjects performing different sets of movements, which is available for researchers worldwide. Energy measures estimated on the wavelet domain is the novel set of features introduced in the sEMG signal analysis community is implemented and compared to already simple features of the time domain. The experimental results show the use of wavelet energies on the wavelet domain can significantly improve the classification challenge.