Motion Discrimination with Minimum Number of Channels Using Low-Cost Myoelectric Sensors in Myoelectric Prosthetic Hands

Abstract

A myoelectric prosthetic hand has the possibility of improving the QoL of patients who have an upper-limb amputation. However, only 2.3% of patients use a myoelectric prosthetic hand in Japan [1]. One of the reason is the price. Most myoelectric prosthetic hand equips many myoelectric sensors and motors to estimate an intention and control hands accurately [2]. In this paper, we discriminates a motion from electromyogram signal recorded by one reasonable myoelectric sensor as a basis of a low-cost and minimum channel control system. Electromyograms for five movements are measured on body surface near four major muscles in the forearm. Feature parameters extracted by electromyogram signals of one muscle is used to classify the motions. The discrimination is performed by feature extracted by one muscle and discriminator of a support vector machine and k-nearest neighbor. We show the discrimination performance of the proposed system using a single channel of the sensor. In addition, the sensor position which maximizes the discrimination performance of our proposal is identified.