Use of a pattern recognition technique to control a multifunctional prosthesis.

Abstract

Various kinds of command source can be used to control an above-elbow prosthesis. But none of them can be used to perform a specified task easily. The research is devoted to investigation of the potential effectiveness of applying the kinematic data of the shoulder joint to control an upper-limb prosthesis. Using these data as input signals, an appropriate signal processing technique, pattern recognition, is utilised to derive control commands. The purpose of the investigation is to use these commands to control a multifunctional prosthesis so that an amputee can perform a few tasks. For testing performance accuracy, a goniometer is worn by a subject with intact arm. It is also interfaced to a digital computer. Next, he is asked to do one of the predefined tasks for which the joint angle trajectories have already been derived and stored. Almost as soon as the shoulder joint angles are sampled and sent to the computer, the program calculates the elbow and wrist angles. These values are compared with actual elbow and wrist angles, which are monitored by the goniometer.