OVPD: Odor-Video Elicited Physiological Signal Database for Emotion Recognition

Abstract

To explore the effects of different stimuli on human emotions, olfaction has been incorporated into stimulation materials that might trigger strong emotions. In this study, we designed a new stimulation experiment and established an odor–video physiological signal database (OVPD) of odor–video stimulation for emotion recognition. The database contained the electroencephalogram (EEG) data of ten subjects acquired when experimenting with video–odor stimuli and video-only stimuli. We selected 32 video clips and ten kinds of odors (i.e., orange, rose, alcohol, water, mint, vinegar, durian, foul, acetic, and myrtle) as elicitation materials. Each subject reported a self-assessment of arousal and valence levels after the experiment ended. By comparing the brain topographies associated with two types of stimuli, olfactory stimuli increased orbitofrontal cortex activity, which is consistent with the finding from the fMRI study. Simultaneously with three types of emotions, we presented the average classification accuracies of different features using the support vector machine (SVM) classifier. The optimal accuracies with or without olfactory stimuli were 99.03% and 97.92%, respectively. The superior results show that the olfactory stimulus can enhance emotional experiences.