Effective extraction of evoked potentials using template cross correlation

Measurements of event related potentials in EEG are contaminated by various sources of noise including artifacts and uncorrelated spontaneous brain activities. A major problem faced during the extraction of evoked potentials is the identification and elimination of bad segments of EEG data contaminated by such sources of noise. We present a new method which automatically identifies and eliminates bad epoch trials not only contaminated by artifacts but also dominated by spontaneous brain activities. This method gives a significantly better estimation of the amplitudes and latencies of the averaged event-related response using only the good epoch trials selected based on template cross correlation as compared with the conventional technique of averaging all epoch trials to obtain the ERP. We illustrate the method by applying it to the extraction of visual evoked potential in the EEG measurements.