Adaptive localization of moving EEG sources using augmented complex tensor factorization

In this paper, an adaptive localization algorithm for moving EEG sources based on tensor factorization is proposed. Moreover, using the augmented complex statistics in tensor factorization enabled us to exploit the full second order information by involving the effect of pseudo-covariance matrix. We simulated EEG signals by using EEG forward solution for moving source dipoles. Then pairing adjacent electrodes to form complex EEG data facilitates the use of cross information. In order to use the recursive least squares tracking (RLST) as an adaptive version of parallel factor (PARAFAC) algorithm, we generated a third order tensor from stacking augmented covariance matrices. The results of this paper compare two windowing schemes of adaptive algorithm with several metrics.