Detection of epilepsy using Hilbert transform and KNN based classifier

Abstract

This work describes a computer aided diagnostic tool for EEG signal classification and analysis. Our main objective is to develop an accurate, automatic and timely classification method for detection of seizures occurring in epileptic patients so that appropriate medical attention can be preemptively provided to the patient The proposed method employs a feature extractor coupled with K-Nearest Neighbor (KNN) classifier. The aforementioned feature extractor is based on the principle of the Hilbert Transform. The KNN classifier is employed for automatic classification of the EEG Signals into two categories viz. healthy subjects and epileptic subjects. The KNN classifier is developed with a training feature vector and then tested with a testing feature vector for binary classification of the EEG signals. Our proposed method has been developed using 5 sets of EEG signals from a publicly available EEG time series database. The average accuracy of our proposed scheme is as high as 91.33%.