Abstract
Autism Spectrum Disorders (ASD) are neurodevelopmental disorders that are associated with characteristic difficulties to express and interpret nonverbal behavior, such as social gaze behavior. The state of the art in diagnosis is the clinical interview that is time intensive for the clinicians and does not take into account any objective measures of behavior. We herewith propose an empirical approach that can potentially support diagnosis based on the assessment of nonverbal behavior in avatar-mediated interactions in virtual environments. In a first study, ASD individuals and a typically developed control group were interacting in dyads. Head motion, and eye gaze of both interlocutors were recorded, replicated to the avatars and displayed to the partner through a distributed virtual environment. The nonverbal behavior of both interaction partners was recorded, and resulting preprocessed data was classified with up to 92.9parcent classification accuracy, with the amount of eye area focus and the average horizontal gaze change being the most relevant features. We expect that such systems could improve the diagnostic assessment on the basis of objective measures of nonverbal behavior.
Users
Please
log in to take part in the discussion (add own reviews or comments).