3D FACIAL VISUAL SPEECH SYSTEM BASED ON MOUTH DEFINITION TABLES IN PORTABLE SETS

摘要

To ensure that 3D facial visual speech system can fast set model of human face and synthesize realistic animation which fit various portable sets for multi-modal interaction, a visual speech system employing Candide-3 model is presented. The system firstly maps the frontal human face photos onto the Candide-3 model to set a human face model, secondly uses the model to establish the mouth definition tables which can adaptively determine the controllable regions of all the mouth feature points and mobility factors of non-feature points controlled by the feature points, therefore the animation can be drove by the mouth feature points with pronunciation. The experimental results of 3D animation using the system demonstrate the animation can reached the level of being acceptable in subjective and objective assessments.