Discriminant Analysis for Epileptic Seizure Detection

Epilepsy is characterized by the sudden and recurrent neuronal firing in the brain. It can be detected by analyzing Electroencephalogram (EEG) of the subject. In this paper, a method of classification of EEG signals into normal and seizure classes is presented. Features based on the statistical distributions were calculated for each frame of EEG signals. After ranking the features using Fisher's discriminant analysis variance, skewness and coefficient of variation (CoV) were found to form the best set of features. Classification was done using linear classifier which showed an accuracy of 96.9%.