A Hybrid Mathematical Model Using DWT and SVM for Epileptic Seizure Classification

This paper illustrates an automatic seizure detection framework that is based on discrete wavelet transforms (DWT), non-linear and statistical features, and support vector machines (SVMs). Electroencephalogram (EEG) signals possess non-linear and rhythmic properties in different frequency bands. Thus, the non-linear features are widely used to advance epileptic seizure detection models and achieve promising results. This research work aims to consider multiple non-linear features so that if the information is missed by one non-linear measure, it can be captured by another. The non-linear features are further combined with the statistical features as statistical features help get better epileptic seizure classification accuracy. All features are calculated on D(2), D(3), D(4), D(5), and A(5) wavelet sub-bands, then combined into a single vector and classified using SVMs. The intended approach’s accomplishment is assessed with respect to terms sensitivity, specificity and accuracy, tested at the University of Bonn and Neurology and Sleep Centre datasets.