Akbar Asgharzadeh - Department of Electrical Engineering, Urmia University, Urmia, Iran
Mehdi Chehel Amirani - Department of Electrical Engineering, Urmia University, Urmia, Iran

One of the most common disorders is epilepsy that approximately 1% of people worldwide are suffering from it. Electroencephalogram (EEG) contains great information about epilepsy, therefore, analysis of EEG can determine the epileptic seizures. In this paper, we propose an efficient method to epileptic seizure detection in EEG signals. After preprocessing and removing frequencies higher than 60 Hz, four-level discrete wavelet transform (DWT) is used to extract five EEG subbands, delta (), theta (), alpha (α), beta (), and gamma (). After that, fractal index is calculated for each subband considering three different methods. In this way, feature vector constructed with 15 features. Finally, these feature are given to support vector classifier (SVM) with different kernels is used to distinguish inter-ictal EEG signal and ictal EEG signals. The results demonstrate that polynomial order of two and Gaussian kernels achieves the highest classification accuracy equals 98.3%. These results demonstrate that proposed method is an efficient method to detect epileptic seizure from EEG signals.

EEG, discrete wavelet transform, fractal index, support vector machine