筋電義手の複合動作と随意的動作を実現する制御系設計

抄録

In this research, we aim to design a control system that realize complex and arbitrary motion for myoelectric prosthesis. In order to enable these operations, we construct a hybrid system to realize complex motion identification by BPNN, and to realize joint angle estimation by MISO-NARX model. The pattern of wrist and arm motion were discerned by BPNN. As a result, rates the identification accuracies are about 95.6%. Moreover, the wrist and arm joint angles were estimated by MISO-NARX model and the estimate errors were 0.167(palm flexion direction), 0.183(ulnar flexion direction), 0.0929(arm turning direction) in mean square error.