Distinction of finger operation for myoelectric prosthetic hand on the basis of surface EMG

Abstract

In this paper, recognition of finger operation via surface EMG is discussed. Finger operation was distinguished into six kinds of motion, and identification of the finger operation was performed using two types of identifiers. One of the identifier is on the basis of combined feedforward neural networks, and the other identifier is on the basis of self-organizing map. A finger operation is carried out in approximately 0.5 seconds for each motion. The recognition accuracy of each identifier was examined for the finger operation, and compared. Finally, the robotic hand for a myoelectric prosthetic hand was controlled according to the recognition result of the finger operation.