Multichannel Feature Extraction and Classification of Epileptic States Using Higher Order Statistics and Complexity Measures

Epilepsy is a brain dysfunction that is characterized by recurrent seizures. An important analysing tool in detection of epilepsy is Electroencephalogram (EEG). The random and non-linear nature of the EEG imposes great difficulty in understanding the pathological process. In this work a multichannel epilepsy detection system is proposed. A feature vector is formed by performing Higher Order Statistics (HOS) and complexity analysis on the signal. Singular Value Decomposition is then used to reduce the dimension of the feature vector. A one-way ANOVA test was performed on the extracted feature vector to select statistically significant singular values [email protected] ( ) 001 . 0 < value p . The selected singular values are used to train the Support vector machine (SVM) based classifier. Here SVM is trained as a patient centric epilepsy classifier as the nature of epilepsy differs between patients. The classification performance of the proposed system is evaluated based on K-fold cross validation technique which showed noteworthy results. Keyword-EEG signal, Poly Spectra, Higher Order Statistics, Singular Value Decomposition, Epilepsy, Complexity Analysis, ANOVA test