Interacting with virtual humans and detection of facial expressions

what makes an avatar more real than real?

Research output: Chapter in Book/Report/Conference proceedingConference proceeding contribution

Abstract

The goal of this project is to investigate how facial expression asymmetry is perceived and reacted to by real humans interacting with virtual humans and whether these reactions are in line with the ones that have been documented in neuropsychological studies of facial expression asymmetry. We recorded short animations of an actor expressively recounting arbitrary passages from children’s stories, using a face-tracking equipment to capture the movement of an actor’s face. The animations were then modified with the three final asymmetry levels (normal [asymmetric], exaggerated [asymmetric] and normalised [symmetric]) and the pairs were combined to allow the participants to rate the difference between each set. To measure user experience in the detection of facial expressions, we used two types of questionnaires: Comparisons and Assessment. In the Comparisons questionnaire, we asked the participants to compare the agents’ facial expressions on four axes: trustworthiness, likability, realism and naturalness. The assessment questionnaire was designed to determine whether the participants read the expressions on the faces of identical models with identical underlying facial movement but differing levels of asymmetry in vastly different ways. We conducted experiments with 52 participants using these questionnaires. We found that the symmetrical faces score the highest rates in all scales vs normal (asymmetrical) faces. The least trusted and liked faces are the ones with exaggerated facial expressions. We observed that the participants tended to use more descriptive words to assess the expressions of virtual humans with asymmetric facial expressions. The results indicate that the participants found the asymmetry more confusing or difficult to interpret when greatly exaggerated.
Original languageEnglish
Title of host publicationProceedings of the IADIS International Conferences Interfaces and Human Computer Interaction 2012 and Game and Entertainment Technologies 2012
EditorsKatherine Blashki
Place of PublicationLisbon, Portugal
PublisherIADIS Press
Pages21-28
Number of pages8
ISBN (Print)9789728939755
Publication statusPublished - 2012
EventIADIS International Conference Game and Entertainment Technologies - Lisbon, Portugal
Duration: 18 Jul 201220 Jul 2012

Conference

ConferenceIADIS International Conference Game and Entertainment Technologies
CityLisbon, Portugal
Period18/07/1220/07/12

    Fingerprint

Keywords

  • Virtual Humans
  • Facial Expressions
  • Face Detection
  • User Experience

Cite this

Kavakli, M. (2012). Interacting with virtual humans and detection of facial expressions: what makes an avatar more real than real? In K. Blashki (Ed.), Proceedings of the IADIS International Conferences Interfaces and Human Computer Interaction 2012 and Game and Entertainment Technologies 2012 (pp. 21-28). Lisbon, Portugal: IADIS Press.