EEG Signal Classification and Segmentation for Automated Epileptic Seizure Detection using SVM Classifier

Epilepsy is an unusual brain activity due to abnormal behavior of neuronal discharge which occurs from concurrent seizures. The electroencephalogram (EEG) is a test that measures the electrical movement in the brain. The main objective of this EEG analysis is to introduce some novel methods for seizure detection and to compare the performance of the Support Vector Machine (SVM) classifier kernels. The 8 statistical features namely mean, standard deviation, median, mode, skewness, kurtosis, maximum and minimum and the 4 Gray Level Co-occurrence Matrix (GLCM) features namely contrast, correlation, energy and homogeneity are extracted from each segment of the EEG signal which are analyzed and classified using two different SVM kernels. The Linear and Radial Basis Function (RBF) kernels are used with the SVM classifier. The performance of the kernels is compared and concluded that the linear kernel is the best choice for this study of seizure detection. Further in this analysis, the parameter sigma in RBF kernel is tuned to the value two and compared with the default sigma value one. It seems that the accuracy is better in classification when the sigma value is tuned.