Using Brain-Computer Interface (BCI) and Artificial Intelligence for EEG Signal Analysis

Abstract

The goal of this paper is to use a brain-computer interface (BCI) and artificial intelligence for EEG signal analysis. This includes answering the question whether it is possible to create a classifier that could correctly recognize emotions from EEG data recorded by simple and cheap equipment. In the paper, we compared the created classifier with the one that was taught on high-quality EEG data, which was recorded with the help of professional-grade equipment. Two experiments were planned. In both experiments, models had to classify three emotions based on EEG data: positive, negative, and neutral. In the first one, a classifier was trained on data from every person in a group and then tested on the data from the same group of people. In the second test, a classifier was trained on data from a group of people and then tested on a person from outside of this group. From the results of the conducted experiments, an appropriate set of conclusions were drawn. It is possible to create a classifier on a lower grade dataset. The best classifier made on the Mindwave dataset had a prediction accuracy of 89%. The classifier made on the lower grade dataset presented worse results than the classifier made on high-grade dataset. The difference in prediction accuracy between both classifiers is 10%. Classifiers tested on one set of people could not accurately predict the emotions of people outside of this group. The conclusions were consistent with related works results.