Perception of emotional expressions in different representations using facial feature points

Facial expression recognition is an enabling technology for affective computing. Many existing facial expression analysis systems rely on automatically tracked facial feature points. Although psychologists have studied emotion perception from manually specified or marker-based point-light displays, no formal study exists on the amount of emotional information conveyed through automatically tracked feature points. We assess the utility of automatically extracted feature points in conveying emotions for posed and naturalistic data and present results from an experiment that compared human raters' judgements of emotional expressions between actual video clips and three automatically generated representations of them. The implications for optimal face representation and creation of realistic animations are discussed.