The WAGL Module: Wearable Actuating Grip Learning Module for Body-Powered Prosthetics

Abstract

Body-powered prosthetic arms are a class of devices that use scapula retraction to pull cables in order to clench a claw/hand attachment. This fully mechanical device is the most commonly used prosthetic arm because of various benefits involving cost, usability, durability, maintenance and more. However, a major drawback of body-powered prostheses is the amount of energy and compensatory motion required to activate it; this contributes to the high rate of prosthetic abandonment. In order to address this, this project proposes the Wearable Actuating Grip Learning (WAGL) Module. This plug-and-play, user-friendly, customizable system can enhance the control of existing body-powered prosthetic arms. This system uses electromyography (EMG) sensors to detect minor muscle movements in order to actuate the body-powered prosthetic, as opposed to scapula retraction. Using machine learning, the user can train a model to detect what muscle movements correspond to gripping, and accordingly actuate a body-powered prosthetic. Experimental results indicate that the framework proposed in this paper can use an on-device, lightweight machine learning model to accurately predict a grip gesture based on limited EMG sensors.