Intelligent Recognition System of Myoelectric Signals of Human Hand Movement

Abstract

This paper presents the recognition of myoelectric signal's algorithm design thru artificial neural network architecture, for the manufacture of a prototype of human hand prosthesis. At the beginning of the project the myoelectric sensor was designed to help capture the signals that correspond to the movement of each finger of a human hand. A database was generated with captured myoelectric signals, which was used for the training of artificial neural networks (ANN), obtaining the weights and bias. The performance of the architecture was evaluated with statistical criteria for the validation of ANN, comparing between simulated data and experimental data. It was found, that the best architecture in this project has 7 neurons in the hidden layer, one in the output layer and 96% correlation coefficient, this architecture is the number 7 in Table 1 which contains a performance report learning algorithm of the different architectures proposed.