Digital filtering in EEG/ERP analysis: Some technical and empirical comparisons

Abstract

A general approach to time domain digital filtering is described, and examples of some filters used in EEG/ERP research are presented. Simulations are reported that evaluate the impact of the relative length of the filter weight series and the signal cycle to be filtered, the span and real-time density of the filter weights, and slow drift across the epoch being filtered. Results indicate that some filters commonly used in the EEG/ERP literature are inadequate. Frequency domain digital filtering is also briefly discussed. The fast Hartley transform, a fast but relatively unknown computational method for frequency domain filtering of ERP/EEG data, is introduced and compared with' time domain filtering. Some practical recommendations are provided. Copyright 1998 Psychonomic Society, Inc.