BodySpeech: A configurable facial and gesture animationsystem for speaking avatars

Abstract

Speaking avatars are present in many HumanComputer Interaction (HCI) applications. Their importancelies in communicative goals which entail interaction withinother avatars in virtual worlds or in marketing where theyhave become useful in customer push strategies. Generatingautomatic and plausible animations from speech cues havebecome a challenge. We present BodySpeech, an automaticsystem to generate gesture and facial animations driven byspeech. Body gestures are aligned with pitch accents andselected based on the strength relation between speech andbody gestures. Concurrently, facial animation is generatedfor lip sync, adding emphatic hints according to intonationstrength. Furthermore, we have implemented a tool foranimators. This tool enables us to modify the detection ofpitch accents and the intonation strength influence on outputanimations, allowing animators to define the activation ofgestural performances.