Variable-Weight Adaptive Filtering as an ERP Pre-processing Technique

Abstract

Motivated by studies about the electrophysiological interference effect due to distinct informative elements of stimulus in response to a simplified variation of the Stroop Test, a previous study proposed a procedure for eye-blink artefact reduction based on adaptive filtering when no eletrooculogram data is available. In that occasion its performance was preliminarily assessed by applying it to real EEG. The present work aims at proceeding with this assessment by means of simulation. The generated signal was supposed to reproduce both EEG, ERP and eye-blink characteristics in four derivations (Fp₁, Fp₂, Cz and Oz). Results show that satisfactory artefact reduction is achieved even when its time occurrence superposes that of the desired ERP. The lowest general RMS error has been achieved using a 6^th order filter and a low adaptation factor. Furthermore, the weight thresholds have minor influence on filter performance when using plausible values. In conclusion, the proposed approach resulted in a considerable reduction of eye-blink artefact, preserving ERP morphology and allowing easy component identification.