Initial assessment of artifact filtering for RSVP Keyboard#x2122;

摘要

RSVP Keyboard™ is an electroencephalography (EEG)-based spelling interface that uses evoked response potential classification with the help of language models. As in any brain computer interface, severe physiological and environmental signal artifacts that affect signal quality in EEG are a detriment to performance. To alleviate the negative effects of such artifacts on RSVP Keyboard™, we implemented a filter that is based on an existing methodology from the literature. Using statistical modeling of pre-recorded EEG that includes three types of artifacts intentionally generated by operators, we perform Monte Carlo simulations of copy-phrase tasks and analyze the effect of artifact filtering on estimated typing performance. The presented results demonstrate an evidence against the usability of the tested method for online artifact reduction applications.