Epileptic seizure detection using ensemble classifier and HOS (Higher Order Statistics)

Abstract

Epilepsy is a neurological disorder including disorders of the nervous system caused by brain damage. In this paper Epileptic seizure detection is done using electroencephalogram signals (EEG). Discrete wavelet transform (DWT) method is a common method to extract four features from the (EEG) signal, and then classify them. To improve and increase the efficiency of extraction of features and to get the best results of classification, we used new method for the extraction of features, namely Higher Order Statistics of Wavelet Packet Decomposition (HOS of WPD). Using this method we get 90 features from every signal, it is then classified using the four classifiers, as follows Artificial Neural Network (ANN), Support Vector Machine (SVM), K-nearest neighbour (KNN), and Naive Bayes. The results of classification are better as compared to DWT.