Evolutionary Feature selection for EEG Signal Classification of Schizophrenic Patients

In this paper, EEG signals of twenty schizophrenic patients and twenty age-matched control subjects are analyzed in order to classify these two groups. From each case, 22 channels of EEG were recorded. Some features including AR parameters, band power and fractal dimension are extracted from the EEG signals. Bidirectional search and plus-L minus-R techniques have been employed to select the more informative channels. After channel selection phase, in order to reduce the feature dimension, genetic algorithm has been applied to select the best features. Finally, Linear discriminant analysis (LDA), multilayer perceptron (MLP), Adaboost classifiers have been chosen to classify these two groups. The final results show %87.00, %96.97 and %86.24 classification accuracy between two groups by LDA, MLP and Adaboost respectively.