Recognition of Lower Limb Movements Baesd on Electromyography (EMG) Texture Maps

To explore the applicability of EMG signal in lower limb movements recognition, this paper proposed a method of constructing the texture image of EMG signal, which realized the intention recognition of lower limb through the way of Convolutional Neural Network (CNN). 10 muscles were selected to collect original signal and 5 actions of stand upright, bent leg lift, straight leg lift, tiptoes and squats were designed for classification in this paper. After Filtering and noise reduction, raw EMG signals were converted to EMG texture maps, as the inputs of CNN, for classifying the above five actions. The experiment results showed that the average recognition accuracy of the five movements of the proposed method has reached 95.5%. At the meantime, the proposed method is more direct and effective, which provides conditions for realizing the real-time perception recognition of lower limb movements in the later stage.