PATTERN SEPARABILITY VISUAL FEEDBACK TO IMPROVE PATTERN RECOGNITION DECODING PERFORMANCE

State-of-the-art myoelectric upper limb prostheses control often utilize pattern recognition (PR) systems that translate electromyograph (EMG) activity to a desired movement. As possible prosthesis movements increase, users have difficulty generating sufficiently separable EMG signals that reliably operate all possible degrees of freedom. Current training regimens attempt to increase the separability of a user's EMG signals through trial-and-error, where a therapist prompts a user to generate EMG signals and provides advice based on the strength and channel distribution of the EMG. In this work, we present a novel visual feedback interface that allows users to observe how their EMG signals affect PR output directly.