The lifted wavelet transform for encephalic signal diagnostic

Several researches and methods have been developed in the aim of efficiently detecting abnormalities in Electroencephalogram (EEG) time series. The aim of this work is to detect a real-time Epileptic seizure. We designed an algorithm which decomposes EEG signals of a database, normal and epileptics, by the lifted wavelet transform (LWT), in order to extract the features. To reduce the time allocated to the classification decision and the parameter space, we applied the principal component analysis (PCA). Among 46 lifted wavelets, the SVM classifier has enabled us to choose, the most appropriate lifted wavelets for this type of application. Our work shows that if we choose the good sampling frequency and good lifted wavelet, with only one vector feature, we can achieved in real time the best classification rate, with a maximum of specificity and a sensitivity.