EEG-Based Emotion Recognition using Beta and Gamma Rhythms With SVM and QBC Classifiers for Emotion-Aware Music Recommendation

Abstract

Conventional music recommendation systems, based on user preferences or genre categories, fall short when it comes to emotionally engaging users according to their personality types. To address this gap, our study investigates EEG-based emotion recognition using Western music in combination with Internet of Things (IoT) technology. Guided by the dimensional model of emotion, we selected three types of Western music fragments designed to evoke emotions induction. EEG signals were then collected and analysed across different brainwave rhythms—theta, alpha, beta, and gamma—using various machine learning classifiers for feature extraction and emotion classification. Our results show that the beta and gamma rhythms produced the highest classification accuracy, with overall averages of 0.842 and 0.841, respectively. Among the classifiers, Support Vector Machine (SVM) outperformed the others, achieving a between-subject accuracy of 95.7% on gamma and 88.2% on beta rhythms, marking a notable improvement over baseline methods. Similarly, the Query-by-Committee (QBC) algorithm achieved up to 90.1% accuracy in gamma and 88.4% in beta rhythms. These findings highlight the potential of SVM and QBC classifiers in improving EEG-based emotion recognition. Interestingly, the most effective EEG features consistently originated from the head region across all participants.