AVDOS - Affective Video Database Online Study Video database for affective research emotionally validated through an online survey

Michal Gnacek, Ifigeneia Mavridou, John Broulidakis, Charles Nduka, Emili Balaguer-Ballester, Theodoros Kostoulas, Ellen Seiss

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

37 Downloads (Pure)

Abstract

High-quality and emotionally validated material is a necessary basis for well-designed experiments in affective cognition and neuroscience research. Videos of existing databases are recorded using and viewed on devices that were current at the time of the design. With the continuous transition to more recent, higher-resolution devices, new databases are needed, and older ones need to be regularly updated. This is particularly true for virtual reality (VR) head-mounted displays (HMD) which, due to the proximity of the eyes to the screen, are extremely prone to low-resolution effects, resulting in unclear images. However, current databases typically feature high intra-variability, combining videos of varied length and quality, which can be problematic when looking at time domain features and physiological signals. In our study, 60 online videos, 30 seconds long were rated by 86 participants in an online survey using a novel class of affective sliders. Mean arousal and valence ratings were calculated for all videos and used for video classification into positive, neutral, and negative categories identified through cluster analysis. Mean ratings showed a parabolic, u-shape curve of arousal and valence present in other similar studies. Two-step clustering has successfully identified 3 clusters we selected and placed 88.3% of videos in the cluster we predicted. Our results suggest that online questionnaires could provide an easier way to update video databases with similar effectiveness as lab-based experiments. Videos and rating table are available for download at: https://gnacek.com/affective-video-database-online-study
Original languageUndefined/Unknown
Title of host publication2022 10th International Conference on Affective Computing and Intelligent Interaction (ACII 2022)
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
Number of pages8
ISBN (Electronic)978-1-6654-5908-2
DOIs
Publication statusPublished - Oct 2022
Externally publishedYes
Event2022 10th International Conference on Affective Computing and Intelligent Interaction (ACII) - Nara, Japan
Duration: 18 Oct 202221 Oct 2022

Publication series

NameInternational Conference on Affective Computing and Intelligent Interaction

Conference

Conference2022 10th International Conference on Affective Computing and Intelligent Interaction (ACII)
Country/TerritoryJapan
CityNara
Period18/10/2221/10/22

Cite this