Real-time Myoelectric Control in a Virtual Environment to Relate Accuracy and Usability

INTRODUCTION Many clinically available, upper-extremity prosthetic limbs provide myoelectric control of a single device, such as a hand, elbow, or wrist. Most commonly, these systems yield control information from myoelectric signal (MES) amplitude [1] or rate of change of MES [2]. Such systems have been beneficial; however, prosthetic users would no doubt find enhanced functionality and usability if they could reliably control more than a single function (or device). Seeking to address this issue, extensive work has gone into developing schemes that provide multifunction myoelectric classification with very high accuracy [3]. However, for all continuous multifunction MES classifiers, no matter how accurate and repeatable, there exists no defined threshold (classification accuracy) of acceptability. This is due, in large part, to the limited availability of prosthetic devices housing multiple electromechanical functions. Described in this paper is a recently developed MES control software tool that incorporates state-of-the-art multifunction control and a multifunction, real-time virtual limb. Twelve subjects have performed a virtual clothes pin functional test [4] yielding results that relate classification accuracy and multifunction device usability. Preliminary results indicate no strong relationship between accuracy and usability scores as determined by this tool.