Time domain analysis of epileptic EEG for seizure detection

Epilepsy is an infirmity which affects the brain causing repeated seizures. An automatic novel method is used for analyzing the EEG signal and for detecting epileptic seizure activity. The proposed method is tested on a publicly available dataset and it uses two time domain features namely line length and energy. Classification algorithms-1) Quadratic discriminant analysis (QdA), 2) K-Nearest Neighbour (KNN) and 3) Linear discriminant analysis (LDA) are used for classifying the EEG signals, and their performance is evaluated by measuring sensitivity, specificity and accuracy. The results of the three classifiers are compared and KNN classifier shows better results than the other two classifiers. An overall accuracy from 94.4% to 100% is achieved by the KNN classifier and the high classification results obtained verified the success of the method.