Surface electromyography(sEMG) is one of the main information sources of human motion detection and has been widely used in the field of well-being of robots. We present a feature extraction method based on wavelet transform power for identifying the movement of human legs. The average power of the active segment in wavelet subspace is used to make up the feature vector according to the frequency domain distribution of the sEMG signal. In order to verify the effectiveness of the proposed method, we design and implement a small portable multi-channel sEMG signal acquisition system, and construct a classifier with support vector machine (SVM) to identify the leg movements. The results of the study show that the method can distinguish four kinds of common actions of the leg, the recognition rate of the same individual can reach more than 95%, and the recognition rate of different individuals can reach 85%, which can be applied to daily rehabilitation training of patients with lower limb movement disorders.
