Prosthetic hand control using motion discrimination from EMG signals

In this report, we improve the motion discrimination method from electromyogram (EMG) for a prosthetic hand and propose prosthetic hand control. In the past, we proved that a motion discrimination method using conic models could discriminate three hand motions without the incorrect discriminations that the elbow motions cause. In this research, to increase discrimination accuracy of motion discrimination using conic models, we propose a feature extraction method using quadratic polynomials. Additionally, because many prosthetic hands using motion discrimination have constant motion speed that can't be controlled, we propose an angular velocity generation method using multiple regression models. We verified these methods by controlling the 3D hand model. In the experiment, the proposed method could discriminate five motions at a rate of above 90 percent without the incorrect discriminations that elbow motions cause. Moreover, the wrist joint angle of the 3D hand model could be controlled by standard variation of 3[deg] or less.