The research on motion recognition based on EMG of residual thigh

Movement pattern recognition is the key to control the intelligent lower limb prostheses. In this paper, surface Electromyographic (EMG) signals from six muscles of the thigh amputee stump were collected. After wavelet de-noised, all the starting and ending time of the effective action was determined by calculating the wavelength of the EMG signal in real time. A variety of time-domain and frequency domain features of the EMG signals were extracted, three movement pattern were recognized based on the Support Vector Machine (SVM) including flat walking, up stairs and down stairs, and the efficiency of identification was improved by feature optimizing. Experimental results show that, the three movement patterns can be classified online by EMG signals from different subjects using the method in this paper, the recognition rate was above 95%, so that just using the stump EMG to recognize the movement intention was proved to be feasible.