Noise cancelation of epileptic interictal EEG data based on generalized eigenvalue decomposition

Denoising is an important preprocessing stage in some Electroencephalography (EEG) applications such as epileptic source localization. In this paper, we propose a new algorithm for denoising the interictal EEG data. The proposed algorithm is based on Generalized Eigenvalue Decomposition of two covariance matrices of the observations. Since one of these matrices is related to the spike durations, we should estimate the occurrence time of the spike peaks and the exact spike durations. To this end, we propose a spike detection algorithm which is based on the available spike detection methods. The comparison of the results of the proposed algorithm with the results of two well-known ICA algorithms (CoM₂ and SOBI) shows that the proposed algorithm denoise data as accurately as these ICA algorithms. The advantage of the proposed technique appears in terms of numerical complexity. The results also show that the proposed algorithm is considerably faster than these ICA algorithms.