EOG Artifacts Removal in EEG Measurements for Affective Interaction

A brain-computer interface (BCI) is a direct link between the brain and a computer. Multi-modal input with BCI forms a promising solution for creating rich gaming experience. Electroencephalography (EEG) measurement is the sole necessary component for a BCI system. EEG signals have the characteristics of large amount, multiple channels and sensitive to noise. The amount of valuable information derived from EEG signals is dependent on both the amount of noises embedded in the original measurement and the algorithms selected for postprocessing. Therefore, artifacts removal in the preprocess step is crucial. Electrooculography (EOG) signals are one of the major artifacts that often appear in EEG measurement. In this paper, we compared two different algorithms (Recursive Least Square (RLS) and Blind Source Separation (BSS)) to investigate their performances on removing EOG artifacts from EEG signals. Results indicate that the performance of RLS algorithm is better than BSS algorithm no matter whether there are any EOG reference signals. For BSS algorithm, the performance is better when EOG reference signals are available. These results show that for a BCI system, EEG reference is often necessary. Performance will be sacrificed if an EEG system cannot have any EOG reference signals.