Classification of Epileptic Seizures using Ensemble Empirical Mode Decomposition and Least Squares Support Vector Machine

Epilepsy is a chronic disorder of the brain in which abrupt disturbance in the brain results in seizures. The occurrence of seizure ranges from a short interval to several hours. Electroencephalogram (EEG) is the most suited tool to diagnose epilepsy. However, currently, neurologists do a manual observation of EEG to diagnose epilepsy and related disorders. Manual observation is a tiresome task and can result in fatigue-related errors in the diagnosis. An EEG based automated diagnosis system is a useful tool in such cases. In this paper, we propose a unique method to classify EEG signals into two classes viz. Seizure and non-seizure using signal Ensemble Empirical Mode Decomposition (EEMD) and Least Squares Support Vector Machine (LSSVM). In this system, EEG signals are decimated using EEMD algorithm. To progress the diagnosis accuracy when compared to human involvement, the LSSVM classifier classifies the signals. The proposed EEMD method is the improved version of EMD attempts to answer the mode-mixing problem of EMD. The feature set comprises of centered correntropy (CenCorrEn) values of decomposed signals. The LSSVM classifier is efficient for this problem as it solves linear equations instead of a quadratic programming problem. The proposed method achieves the highest classification accuracy of 94.7%on Bonn EEG dataset.