Classification of Sensorimotor Rhythms Based on Multi-layer Perceptron Neural Networks

Abstract

Sensorimotor rhythms are represented by mu rhythm with 8-12Hz frequency band and beta rhythm with the 12-30Hz frequency range. The movement or preparation of the movement is typically accompanied by a decrease of the mu and beta rhythms, especially in the contralateral area of the movement, which is a piece of very important knowledge for the implementation of a brain computer interface. The EEG signal was recorded using 8 active electrodes placed in the motor areas of the scalp. Features extraction was performed by decomposing the original signal in subcomponents signal with the frequency band in the interest range using multiresolution wavelet analysis based Daubechies wavelets. Multi-layer perceptron neural networks (MLP-NN) method is utilized for the classification of the features. This classifier performs well, 95.45% was the maximum classification rate for the subjects involved in the study. The superiority of the classifier MLP-NN was sustained also by The Friedman Two-way Analysis of Variance (ANOVA) by Ranks Test.