Automated detection of epileptic seizures using DWT based features and SVM classifier

Automated detection of epileptic seizures has gained significant attention in the recent decades. This is due to the fact that it helps neurologist to take timely decision and reduces the manual intervention of assessing electroencephalogram (EEG) recordings. Therefore, in this study, the discrete wavelet transform (DWT) features based automated detection of epileptic seizures has been proposed. EEG signal was decomposed using DWT with Haar wavelet and eleven feature were extracted from each sub-band. The extracted features in each sub-band were classified using support vector machine classifier with 10-fold cross-validation. Classification results showed the highest sensitivity, specificity, accuracy and F measure of 97.37%, 98.88%, 98.06%, and 97.84 % respectively using the Ramaiah Memorial College and Hospitals database. Similarly, the highest sensitivity, specificity, accuracy and F measure of 98.90%, 99.62%, 99.18%, 99.17% were achieved respectively using University of Bonn database. The experimental results show that the proposed algorithm is well suited for real-time detection of epileptic seizures.