Robust EEG preprocessing for dependence-based condition discrimination

Abstract

This paper addresses the robustness of the filtering schemes in processing high resolution electroencephalogram (EEG) data in the context of discriminating two stimuli flickering at a given frequency. The raw data consists of recordings from a 128-channel HydroCell GSN where the subject was visually stimulated with two images flickering at 17.5 Hz, representing two distinct conditions, referred to as Face and Mock. These signals were then passed through a band pass filter to only capture the modulation at the flickering frequency, and a connectivity analysis was performed on the filtered signal using generalized measure of association, to observe if the network connectivity changes from one stimulus to the other. In this paper, we investigate the effect of the bandpass filter on the discriminability of the stimuli over different filter orders and quality factors. We observe that the network connectivity is stable over a significant range of parameter values of the filter, thus establishing the desired robustness.