Epileptic detection based on EMD and sparse representation in clinic eeg

Abstract

The sparse representation has gained considerable attention in pattern classification recently. A new method is designed to detect seizure EEG signals by empirical mode decomposition (EMD) and sparse representation. First of all, the EMD method is used to process EEG signals for generating IMFs. Then frequency features of IMFs are extracted as the dictionary in sparse representation based classification (SRC) scheme to realize reduction of the data dimension and calculation cost. Finally, the KSVD is exploited to optimize dictionary. It has been showed that the algorithm can well detect the seizure EEG signals and the accuracy is up to 99%. Furthermore, the quick speed makes the method practical for the treatment of epilepsy in practice.