For a myoelectric prosthetic robot hand to be useful in everyday life situations, an accurate mapping of the muscle EMG signals to each finger motion is required. We propose a deep learning-based method for mapping the surface EMG signals to the hand gesture recognition. In our method, the EMG row data are used as input of a Convolution Neural Network (CNN) and the features of the EMG data are generated in the process of learning. The results show a good performance of CNN for most of 53 considered gestures. In addition, the trained CNN performed well also in real time robot control.

Ikeda Komi