Videorealistic facial animation for speech-based interfaces

Abstract

This thesis explores the use of computer-generated, videorealistic facial animation (avatars) in speech-based interfaces to understand whether the use of such animations enhances the end user's experience. Research in spoken dialog systems is a robust area that has now permeated everyday life; most notably with spoken telephone dialog systems. Over the past decade, research with videorealistic animations, both photorealistic and non-photorealistic, has reached the point where there is little discernible difference between the mouth movements of videorealistic animations and the mouth movements of actual humans. Because of the minute differences between the two, videorealistic speech animations are an ideal candidate to use in dialog systems. This thesis presents two videorealistic facial animation systems: a web-based system and a real-time system.