Optimized Recognition Method of Surface EMG Signals Multi- Parameters Based on Different Lower Limb Motion Velocity

Surface electromyography (sEMG) signal is one of major neural control signal sources for powered prostheses and rehabilitation robots. We focus on exploring the performance of the classification of surface electromyography signals based on different speeds for assisting control of lower limb prosthesis. In order to get the most representative feature of the sEMG signals, several features are extracted based on the integrated electromyography; root mean square; median frequency; mean power frequency and fractal domains. The surface electromyography signals are acquired from thirteen volunteers with three different speeds. The back-propagation neural network and support vector machine are used to recognize movement pattern of the different speeds. The results presented that the features in time-domain have a better representativeness of the surface electromyography signal and the back-propagation neural network have a higher rate of the identification.