Personal Authentication and Hand Motion Recognition based on Wrist EMG Analysis by a Convolutional Neural Network

Abstract

Recent years, EMG has attracted much attention as a tool of human interface. In hand motion recognition and personal authentication using wrist EMG, we have obtained good results. However, there has been no way to establish them at the same time. Therefore, in this paper we measure EMG by attaching dry type sensors to wrist, and carry out hand motion recognition and personal authentication. The conventional method used EMG of movement Japanese Janken. We use a multi-input and multi-output model of a Convolutional Neural Network (CNN). The average accuracy of hand motion recognition is 94.5%. The average accuracy of personal authentication is 94.6%. In the conventional method, personal authentication was classified into two classes. However, we carry out multi-class classification in the proposed method. In feature extraction, we obtain 128×8 input data from the measuring unit. Then, a filter size of the convolution layers is 3×3. CNN does not contain pooling layers in this paper. In the proposed method, the average accuracy of hand motion recognition is 94.6%. The average accuracy of personal authentication is 95.0%.