Automatic Detection of Epileptic Seizures Using Feature Extraction Methods from EEG and its Sub bands

In this paper classification of Electroencephalogram (EEG) signals of 21 epileptic patients for seizure detection is presented. Feature extraction from EEG is done using chaoticfeatures i.e. Fractal Dimension and Approximate Entropy and also by applying Discrete Wavelet Transform, these chaoticfeatures and also some statistical features were extracted from EEG sub-bands: Delta (<4Hz), theta (4-8Hz), alpha (8-12 Hz), beta (13-30 Hz) and gamma (>30 Hz). This methodology isused to classify three classes: pre-ictal (just before the seizure), ictal (during a seizure) and post-ictal (after the seizure).Calculating descriptive statistics of each feature using ANOVA shows that all of them were separating three states from eachother. EEG Channel Selection based on Mutual Information is used for improving classification accuracy, Four classifiers were evaluated and compared including Adaptive Neuro-FuzzyInference System (ANFIS), Artificial Neural Network (ANN), KNearest Neighbourhood (KNN) and Radial Basis Function (RBF)

کلیدواژه ها:

DWT ، EEG sub band ، EEG classification ، epilepsy ، wavelet-chaos