The Impact of Odor Category Similarity on Multimedia Experience

Abstract

Although we have multiple senses, multimedia mainly targets vision and olfaction. To expand the senses impacted by multimedia, olfactory stimulation has been used to enhance the sense of reality. Odors are primarily matched with objects in scenes. However, it is impractical to select all odors that match all objects in a scene and offer them to viewers. As an alternative, offering a single odor in a category as representative of other odors belonging to that category has been suggested. However, it is unclear whether viewers' responses to videos with multiple odors (e.g., rose, lavender, and lily) from a category (e.g., flowers) are comparable. Therefore, we studied whether odors belonging to a given category could be similar in behavioral congruency and in the five frequency bands (delta, theta, alpha, beta, and gamma) of electroencephalogram (EEG) data collected while viewers watched videos. We conducted questionnaires and EEG experiments to understand the effects of similar odors belonging to categories. Our results showed that similar odors in a specific odor category were more congruent with videos than those in different odor categories. In our EEG data, the delta and theta bands were mainly clustered when odors were offered to viewers in similar categories. The theta band is known to be primarily related to the neural signals of odor information. Our studies showed that choosing odors based on odor categories in multimedia can be feasible.

Keywords: Cluster analysis; EEG; Machine learning; Multimedia; Olfactory perception.

Fig. 1

Stimuli and experimental procedure. (a) In this experiment, two videos and four odors were utilized. There were three distinct groups based on the odors featured in the videos. "Group1": FO1 is in the FV, and CO1 is in the CV. "Group2": FO2 in the FV, CO2 in the CV. "Group3": CO1 is in the FV, FO1 is in the CV. (b) Procedures employed in this study. (c) The experimental procedure in the EEG recording test comprised three steps. The first step was a 10-second baseline during which participants acclimated to the experimental situation. The second step was the test, where one of the three odors was presented to participants for 5 seconds at the beginning of the videos, depending on the assigned groups. The final step was a 30-second rest before the subsequent round of the experiment.

Fig. 2

Four odors were grouped into two categories. (a) Cumulative scores of four odors across 19 odor categories (citrus, fruit, coffee, flower, nut, green, tree, root, herb, sweet, powdery, creamy, soap, spicy, musk, mint, watery, tropical fruits, and others). (b) Results of the discrimination test for the four odors. The Y-axis represents the accuracy of correct responses, with the red line indicating the level of random accuracy in this test. Wilcoxon signed rank test was performed for this test. ***p value<0.001.

Fig. 3

Significantly higher congruency with videos in similar odor categories than different odor categories. The Y-axis indicates the percentage of congruency between the odors and the video, while the X-axis indicates the odors corresponding to the videos. (a) Results of congruency between the three odors and the FV. Significantly higher congruency was observed in FO2 compared to CO1. (b) Results of congruency between the CV and three odors. Congruency in CO1 and CO2 was significantly higher than congruency in FO1. *p value<0.05, **p value<0.01, ***p value<0.001.

Fig. 4

Representative event-related spectral perturbation (ERSP) data of the videos with the odors depending on the categories of odors. X-axis is the time and Y-axis is the frequency. (a) Representative ERSP data of the FV and three odors. (b) Representative ERSP data of the CV and three odors.