Partial Brain Model For real-time classification of RGB visual stimuli: A brain mapping approach to BCI

Abstract

Brain-computer interface (BCI) applications are characterized by real-time feature extraction and classification. Thus, brain activity reconstruction is not generally performed due to the long computation times required to estimate brain activity. Here, we present a method for applying brain mapping solutions to BCI applications, based on a reduced model of the brain covering the occipital region using only four local electrodes (P3, P4, O1, and O2) for the classification of red, green, and blue (RGB) visual stimuli. We obtained a classification accuracy of 100% within feasible estimation times for BCI applications using an event-related potential (ERP) dataset. We first validated the occipital-zone model with a test using synthetic EEG data to evaluate its performance for local brain mapping with a small number of electrodes. In a second test, we used the validated partial model to map the occipital and part of the parietal lobes for RGB classification with outstanding results.