A new method for the detection of epilepsy and epileptic seizures based on the variance of EEG signals and its derivatives with a simple kernel trick

In this paper, we propose a new method for the automatic diagnosis of epilepsy via encephalographic signals (EEG). Our objective is the detection of epilepsy and epileptic seizures through EEG of healthy subjects (H), epileptic subject (E) and epileptic subject during seizures (S). Two novelties are deliberated in this paper. In the first method, we have exploited EEG and its derivatives, which gives significant results from calculations of just three features, the variances of the signals and its first and second derivatives. In the second one, we have used a kernel trick that allows an implicit redescription of the extracted features, by the conversion of the nonlinear problem to linear space, which ultimately facilitates the classification step and gives reliable result in fast running time. The experimental test via the Bonn EEG dataset proves the efficiency of the proposed method, an accuracy of 100 % is achieved in seizures detection problem and of 99.8 % in epilepsy detection problem, moreover for the differentiation of three cases 99.85 % of accuracy was achieved.