Classification of Emotions under Multiple Olfactory Stimuli Using EEG Signals and Machine Learning

Abstract

Accurately predicting emotional states is crucial in various fields, including cognitive sciences, artificial intelligence, human-computer interaction, and neuromarketing. In particular, the development of emotion-focused technologies requires the analysis of emotional responses with objective biological data. In this study, the emotional responses of individuals to different odor stimuli were examined using electroencephalogram (EEG) signals and classified using machine learning methods. A total of 46 participants were exposed to an odorless condition and four different odor stimuli (cinnamon, citrus, green tea, and lavender). After each exposure, participants filled out self-report questionnaires based on the valence-arousal model. Emotional states were predicted using time-domain features obtained from EEG signals, and various machine learning algorithms were applied for classification. The results show that EEG-based approaches can classify emotional responses with high accuracy, with lavender being the odor that created the most potent effect, achieving an accuracy rate of 80.14%. This study demonstrates that emotion analysis using EEG signals combined with subjective assessment has significant potential in areas such as neuromarketing and therapeutic applications.