A Novel SVM and K-NN Classifier Based Machine Learning Technique for Epileptic Seizure Detection

Abstract

An EEG signal is used for capturing the signals from the brain, which helps in localization of epileptogenic region, thereby which plays a vital role for a successful surgery. The focal and non-focal signals are obtained from the epileptogenic region and normal region respectively. The localization of epileptic seizure with the help of focal signal is necessary while detecting seizures. Hence, the present article provides detailed analysis of EEG signals. The Focal and Non-focal signals are decomposed using EMD-DWT. A combination of EMD-DWT decomposition method in accordance with log-energy entropy gives an efficient accuracy in comparison to other entropy in differentiating the Focal from Non-focal signals. The extracted features are subjected to SVM and KNN classifiers whose performance will be calculated and verified with respect to accuracy, sensitivity and specificity. At the end, it will be shown that KNN produces the highest accuracy when compared to SVM classifier.