Identification of Epileptic Seizure in EEG Signals Using DWT and ANN

Abstract

Epileptic seizure is the abnormal electrical activity of EEG signal. As EEG signal is a non-stationary signal, DWT is a powerful tool to interpret such kind of signals both in time and frequency domains. In this study, DWT is performed using the number of mother wavelets to measure their classification performance. Characteristic features extraction from the sub-bands, as a consequence of DWT, plays the key role to classify the signal accurately. A new feature set, consists of four nonlinear statistical features obtained from each subbands, has been fed to the input of ANN. To evaluate the performance of our study a famous publicly available dataset is used. This study is conducted on four classification problems mixture of normal and seizure segments. 100%, 99.33%, 97.33% and 98.4% classification accuracy are achieved for Class 1, Class 2, Class 3 and Class 4 classification problems, respectively.