MATLAB/simulink-supported EMG classification on the raspberry pi

Abstract

Patient satisfaction with body-powered myoelectric upper limb prostheses is limited, often resulting in device abandonment. Multifunctional hand prostheses are one potential solution to increase patient acceptance. These require sophisticated control schemes like patternrecognition- based approaches involving classification of myoelectric signals (MES). To allow fast and flexible evaluation of prosthesis control approaches, a prototyping environment based on the Raspberry Pi and MATLAB/Simulink was created. It supports commonly applied features like RMS and zero crossings as well as classification methods like Naive Bayesian and Support Vector Machine classifiers. After classifier training with a custom MATLAB application, MES can be classified in real-time and the results employed for prosthesis actuation. The setup was tested with five participants for controlling a Michelangelo Hand. Over 90% of movements were correctly identified for three classes from two channel EMG data.