Brainwave Classification of Visual Stimuli Based on Low Cost EEG Spectrogram Using DenseNet

Abstract

The paper presents the classification of brainwave response based on low-cost Electroencephalograph (EEG) spectrogram. The low-cost EEG can record EEG data, but it performs significantly worse than the medical device. The purpose of our research is to understand the way of images are perceived by people based on their brain waves. Based on the EEG, we investigate two paradigms in mind-brain research that is: the valence-arousal plane and the universal symbol. The experiments are done using two paradigms to generate EEG data based on visual stimuli and convert it into spectrogram. For brainwave response classification based on EEG spectrogram, we use the recent deep learning architecture that is DenseNet. Furthermore, we compare the training performance between two brain-mind paradigms. The experiment results show that: (i) the EEG signal classification of visual stimuli based on valence-arousal model framework is not reliable, (ii) the universal symbol paradigm gives a promising result in generating visual stimuli for brainwave response classification based on low-cost EEG data.