Investigating critical brain area for EEG-based binocular color fusion and rivalry with EEGNet.

Abstract

Binocular color fusion and rivalry are two specific phenomena in binocular vision, which could be used as experimental tools to study how the brain processes conflicting information. There is a lack of objective evaluation indexes to distinguish the fusion or rivalry for dichoptic color. This paper introduced EEGNet to construct an EEG-based model for binocular color fusion and rivalry classification. We developed an EEG dataset from 10 subjects. By dividing the EEG data from five different brain areas to train the corresponding models, experimental results showed that: (1) the brain area represented by the back area had a large difference on EEG signals, the accuracy of model reached the highest of 81.98%, and more channels decreased the model performance; (2) there was a large effect of inter-subject variability, and the EEG-based recognition is still a very challenge across subjects; and (3) the statistics of EEG data are relatively stationary at different time for the same individual, the EEG-based recognition is highly reproducible for an individual. The critical channels for EEG-based binocular color fusion and rivalry could be meaningful for developing the brain computer interfaces (BCIs) based on color-related visual evoked potential (CVEP).