EMG-Based Interface Using Machine Learning

Abstract

This paper presents an EMG (electromyogram)-based input interface using machine learning for people with physical disabilities of the extremities. We have developed a virtual hand that can be operated in virtual environment using EMG signals. In this paper, we performed a lifting object task and box and block test task with the virtual hand. From the experimental results of the lifting object tasks, it was confirmed that six wrist joint movements were classified, and that an experimental subject appropriately lifted objects with the virtual hand in the virtual space. In the box and block tests task, it was confirmed that he moved block(s) to the opposite side of the box 9 times within 60 sec.