Detection of epilepsy using discrete cosine harmonic wavelet transform-based features and neural network classifier

Abstract

Epilepsy is a neurological disorder caused by the sudden hyper activity in certain parts of the brain. Electroencephalogram (EEG) is the commonly used cost effective modality for the detection of epilepsy. This paper presents a method to detect epilepsy using discrete cosine harmonic wavelet transform (DCHWT) and a neural network classifier. DCHWT is a harmonic wavelet transform (HWT) based on discrete cosine transform (DCT), which is proved to be a spectral estimation technique with reduced bias is used in this work. The proposed method involves decomposition of EEG signals into DCHWT sub-bands, extraction of features from sub-bands and classification using an artificial neural network (ANN) classifier. The main focus of this study is the automatic detection of epilepsy from interictal EEG. This is still a challenge to the researchers as interictal EEG looks like normal EEG which makes the detection difficult. The proposed method is giving classification accuracy of 93.33% to 100% for various classes.