EMG Pattern Recognition System Based on Neural Networks

In this document we present a methodology for movement pattern recognition from arm-forearm myoelectric signals, starting off from the design and implementation of an electromyography (EMG) instrumentation system, considering the Surface EMG for the Non Invasive Assessment of Muscles (SENIAM) rules. Signal processing and characterization techniques were applied using the pass-band Butter worth digital filter and fast Fourier transform (FFT). Artificial neural networks (ANN) such as back propagation and radial basis function (RBF) were used for the pattern recognition or classification of the EMG signals. The best results were obtained using the RBF ANN, achieving an average accuracy of 98%.