筋電位に基づく個人認識の精度向上

―筋電位に基づいた個人認識システムの構築

抄録

In this research, we aim to design a highly accurate personal recognition system using myoelectric potential. A general method of personal recognition based on EMG utilizes muscle potential during certain gestures. The key idea of the general method of personal recognition based on EMG signals is to construct a discriminator that can identify the gesture from the EMG signals generated when a certain gesture is performed. The principle of the discriminator is that when a series of gestures is performed, the discriminator identifies the correct gesture if the EMG signals are generated by the true user, but does not from the EMG signals generated by other users, even if they perform the same gesture. In previous studies, there has been a problem of high misidentification rate, i.e., the system does not recognize the true user or misidentifies other users as the true user. Therefore, this study strives to improve the accuracy of personal regcognition systems based on EMG by using deeplearning to find personal features, and based on features to construct a discriminator that can identify the gesture from the EMG signals generated.