Cost-effective system for the classification of muscular intent using surface electromyography and artificial neural networks

This paper presents the implementation of a system to classify muscular intent. A neural network is used for this purpose. After skin preparation, feature extraction, network training and real-time testing, an average overall classification accuracy of 93.3% over three possible gestures was obtained. Ultimately, the results obtained speak to the suitability of an Arduino-based system for the acquisition and decoding of muscular intent. This result is indicative of the potential of the Arduino microcontroller in this application, to provide effective performance at a far lower price-point than its competition.