CLASSIFICATION OF ICTAL AND INTERICTAL EEG SIGNALS

摘要

An electroencephalogram (EEG) is a graphical record ofongoing electrical activity produced by firing of neuronsof the human brain due to internal and/or external stimuli.Feature extraction and classification of the EEG signalsare used for diagnosis the epileptic seizure (i.e., physicalchanges in behaviour that occur due to abnormal electricalactivity in the brain). Classification of Ictal (i.e., seizureperiod) and Interictal (i.e., interval between seizures) EEGsignals is very important for the treatment and precautionof an epileptic patient. However, the classificationaccuracy of Ictal and Interictal EEG signals is not atsatisfactory level due to their non-abruptness phenomenonusing the existing seizure and non-seizure classificationmethods. Moreover, the features of Ictal and Interictalsignals are not consistence in different locations for anepileptic period. In this paper we present new approachesfor features extraction of Ictal and Interictal using varioustransformations such as discrete cosine transformation(DCT), DCT-discrete wavelet transformation, andsingular value decomposition. The least square supportvector machine is applied on the features forclassifications. Results demonstrate that our proposedmethods outperform the existing state-of-the-art methodin terms of classification accuracy for the largebenchmark dataset in different brain locations.