Tracking of nonstationary EEG with Kalman smoother approach

An adaptive autoregressive moving average (ARMA) modelling of nonstationary EEG by means of Kalman smoother is presented. The main advantage of the Kalman smoother approach compared to other adaptive algorithms such as LMS or RLS is that the tracking lag can be avoided. This advantage is clearly presented with simulations. Kalman smoother is also applied to tracking of alpha band characteristics of real EEG during an eyes open/closed test. The observed tracking ability of Kalman smoother, compared to other methods considered, seemed to be better. Keywords— Nonstationary EEG, Kalman smoother, adaptive algorithms