Classification of Hand Movements from EMG Signals using Optimized MLP

摘要

The emergence of artificial neural networks (ANNs) caused a great revolution in the knowledge area of pattern-classification and contributed to the fast development of intelligent systems. For an ANN to be effective, its topology parameters must be set up, such as learning rate, number of layers, and number of neurons. Therefore, optimization algorithms can be used to produce more efficient ANNs. In this paper, we propose a genetic algorithm (GA) to better estimate parameters of an ANN applied to the classification of the type, strength, and orientation of different hand movements from electromyography signals. Acquired data signals were first processed with the Wavelet Transform (WT) to extract the most important features. These features were presented to the optimized hybrid intelligent system (HIS) composed of ANNs and the optimization GA. Classification results are promising, presenting accuracy rates above 90% and reliability rate of 98%. These results support the use of HIS in practical applications such as smart prostheses and automatic medical evaluations.