ANN-based classification of EEG signals using the average power based on rectangle approximation window

In this study, EEG signals were classified by using the average powers extracted by means of the rectangle approximation window based average power method from the power spectral densities of frequency sub-bands of the signals and two different artificial neural networks (ANNs) which are adaptive neuro-fuzzy inference system (ANFIS) and multilayer perceptron neural network (MLPNN). In order to evaluate their performances together the proposed approach, four different experiments were implemented by using different mixtures of classes. The experiments showed that both classifiers with the proposed approach resulted in satisfactory classification accuracy rates, although the success of MLPNN classifier was a little better than the other. Streszczenie. W artykule zaprezentowano klasyfikacje sygnału EEG przy wykorzystanoiu widma gęstości mocy w podzakresach częstotliwości oraz sieci neuronowych: adaptacyjnego system neuro-fuzzy ANFIS oraz wielowarstwowego perceptronu MLPNN. (Klasyfikacja sygnału EEG wykorzystująca sieci neuronowe oraz uśrednioną moc w oknie prostokątnym)