Recent Applications of Parametric Modeling to EEG Signals Analysis

In this paper, recent applications of autoregressive (AR) and adaptive autoregressive (TVAR) models to EEG signals for detection of epileptic seizures are addressed. First of all, AR/TVAR models based the complexity measure with the order of AR model and the spectrum estimation with online AR model are introduced and employed to analyze the EEG signals with epileptic seizures. Then, three new applications of AR and TVAR models to EEG signals are addressed in detail. The first method is to combine a TVAR model and a dynamic linear model to decompose an EEG signal for revealing the latent structure of the EEG signal. The second method is to combine the information retrieve and AR model to discover the dynamic characteristic, damping time, of an EEG signal. The synchronization measure of multi-channel EEG signals based on AR model is the third method. The test results of real EEG signals showed that the decomposition, the damping time and the synchronization measure of the EEG signals could discover the characteristics of epileptic seizures and help to diagnose the epilepsy patents.