The effects of different wavelet degrees on epileptic seizure detection from EEG signals

Abstract

In this study, EEG records taken from healthy people with eyes open and eyes closed, EEG records taken from epileptic patients at the time of seizure and out of seizure were classified using Naive Bayes, K-Nearest Neighbor and Artificial Neural Networks methods. Feature vectors are obtained by using Daubechies wavelet transforms with different degrees and their effect on the classification success is examined. When the results are evaluated, it is determined that Artificial Neural Networks algorithm is the most successful method using db3 and db5 wavelet coefficients as feature vector. Based on the results obtained in this study, it is thought that the recommended methods will help the experts to decide on the epileptic seizure.