Epileptic Seizure Detection in EEG Signal using EMD and Entropy

: Epilepsy is a disease caused by abnormal electrical activity in the brain. One of the techniques for diagnosing epilepsy is by analyzing electroencephalogram (EEG) signals. Various techniques were developed by researchers to analyze epileptic seizure on EEG signals. Because of the nonlinear, non-Gaussian, and nonstationary nature of EEG signals, methods such as empirical mode decomposition (EMD) are often used for analysis on EEG signals. The intrinsic mode function (IMF) of the EMD is believed to provide different information for normal EEG and seizure signals. Some features are taken from the IMF such as statistical features and spectral features. One of the differences between normal signals and abnormal signals is signal complexity where one of the metrics for measuring them is entropy. Several research used entropy combined with first and second order statistical features. In this study, only one entropy feature used to characterize each IMF produced from EMD for the classification of epileptic seizure EEG. Entropie used were Shannon entropy (ShEN), spectral entropy (SE), Renyi entropy (RE), and permutation entropy (PE). The highest accuracy produced by RE in eight IMF uses quadratic support vector machine (SVM) as the classifier. The accuracy of 97.3% with sensitivity of 97% and specificity of 99.75 % was achieved for classification in three data classes. The developed method is able to produce high accuracy with a relatively small number of features. type first of of