Epileptic seizure detection using PCA on wavelet subbands

The detection and classification of epileptic seizures using the Electroencephalography (EEG) signal has been an active field of research from past few decades. EEG is a non-stationary signal that represents electrical activity along the scalp containing very useful information about normal or epileptic brain states. In this paper, principal component analysis is performed on the wavelets subbands of normal & epileptic signals using discrete wavelet transform. This method is applied to two different groups of EEG signals, i.e., (1) Healthy states (2) Epileptic states, during a seizure (ictal EEG). The features extracted from the principal components that are evaluated from the wavelet subbands, differentiate between these two states. Further, t-student statistical distribution is applied to determine the measure of distinguishing between different subjects. The method of principal component analysis on wavelet subbands can discriminate between ictal & non-ictal subjects with 99.99% pvalue (eye open) and 99.96% p-value (eye closed) using the delta subband. The results presented here are much better than the results of previous researches.