Ocular Artifact Minimization by Adaptive Filtering

Abstract

The problem of real-time ocular or eye artifact correction is addressed in this paper based on the framework of the general adaptive interference canceler, wherein the EOG signals are used as the reference signal. Adaptive algorithms such as LMS, recursive LS, or exponentially weighted LS can be used to update the coefficients of the adaptive filter. The major problem associated with an adaptive eye artifact canceler is found to be the unwanted correlations between the desired and reference signals. This is especially problematic when slow cognitive potentials or slow head or body movement artifacts coexist with eye artifacts in the recorded EEG. Undesired correlations can result in over-correction of ocular artifacts if a standard adaptive filter is used. We tackle this problem by taking into account a priori information regarding the ocular artifacts, that is, the spatietemporal statistics of the transmission coefficients. This strategy yields an adaptive artifact canceler combined with leakage and signal subspace enhancement.