AN EXTRACTION AND CLASSIFICATION BASED ON EMD AND LSSVM OF EPILEPTIC EEG

Abstract

The epilepsy EEG signal with nonstationary and nonlinear characteristics is a typical electroencephalographic (EEG), which have been solved by the proposed method of using least squares support vector machine (LSSVM) as the classifiers and empirical mode decomposition (EMD) as the feature extraction methods to achieve epileptic seizure detection with signal analysis and processing. In this proposed method, EMD is used to select three intrinsic mode functions (IMF) with high correlation to replace the original signals, which has been employed to solve the nonlinear and nonstationary problems in feature extraction and infusion, and then, the feature can be employed to feed to the recognition engine named LSSVM, and its parameters is optimized by particle swarm optimization (PSO). The study uses publicly available EEG database from the University of Bonn (UoB), and there are 7960 EEG segments in the complete dataset, among which are nine recognition problems marked as Z-N, Z-F, Z-S, O-N, O-F,O-S N-S F-S and Z-O-N-F-S, the average classification accuracy of Z-N, Z-F, Z-S, O-N, O-F, O-S, N-S and F-S can be generally obtained as highly as 90%, the Z-O-N-F-S training set and test set classification accuracy are 98.8% and 88%, which had been used to verify the effectiveness and robustness of this proposed method on feature extraction, To the best of our knowledge, the excellent performance of the proposed method has shown that this method can be employed to track the patient’s healthy state and monitor the moment of epilepsy seizure.