Simultaneous myoelectric pattern recognition using biopatrec platform for hand prosthesis

Abstract

Currently, commercial hand prosthesis use a sequential control mesh, which makes the movement of the prosthesis counter-intuitive and clunky, often dependent on external sensors for movement execution. Pattern recognition is a method that has been developed to address these limitations. Unlike traditional strategies, pattern recognition is based on the idea that learning is done by a classification software. For that, the subject can use the natural contractions of the movement that one wishes to control. The software identifies the muscle pattern and classifies it as a target movement. Then, it will recognize the pattern the next time it is generated and create the intended prosthetic movement. In this work were proposed the combination of several methods for feature extraction together with feature selection, applying multilayer perceptron network (MLP) to recognize the motor pattern, using the BioPatRec platform (Ortiz-Catalan et al. in Source Code Biol Med, 2013, [1]). BioPatRec is an open source platform, that allows the implementation and test of several algorithms in the fields of signal processing, feature extraction and selection, pattern recognition and real-time control. The experimental results showed that the proposed features could achieve an average classification accuracy of 97.88%, which was 4.54% higher than the analysis without the features proposed in this work. The results suggest that the new features and the addition of featuring selection have the potential for the use with a myoelectric prosthesis with simultaneous control.