부산대학교 도서관

EEG emotion recognition based on audio and video stimulation has made some progress, but there are few researches on EEG emotion recognition based on video game stimulation. Video games, due to their high interactivity, can be an effective emotion recognition tool. Therefore, this paper uses four types of games that represent 4 different emotions (high arousal positive valence, high arousal negative valence, low arousal positive valence, low arousal negative valence) and collects the EEG signals of 15 subjects during the game. For EEG feature extraction, this paper extracted five functional brain network features, namely Pearson correlation coefficient, phase-lock value, phase lag index, Granger causality and mutual information. The results show that the combination of CNN model and the phase-locked value feature of beta band has the best classification effect, and the average accuracy of each subject is reached 93.86%. In addition, it is found through visualization that the emotion-generating regions based on video games are mainly concentrated in the frontal and occipital lobes of the brain.