Research On Parameterized Face Expression Simulation Method Based On 3D Grid Model

Face expression simulation animation is one of the most chanllening problems in Computer Graphics. It can be widely used in many application areas, such as computer animation industry, game industry, teleconferencing, and agents and avatars. In recent years, it has been the research focus.The paper gives a comprehensive survey and analysis of existing face expression simulation, and describe the mothods of current face expression animation. By systematically analyzing relevant algorithms, we present our novel methods for face expression simulation. We employs standardized model to describe face expression feature, and improves some technology of some aspects such as organization , extration and driving of expression parameters. In addition, we implement a face expression simulation system. The main research work is as followed:(1) Select advanced Candide-3 model as standard face grid model. Referring to MPEG-4 face standard and facial action coding system, modify and amend the defects of Candide-3 model.(2) Extract expression paremeters based on Cohn-Kanade face expression database. According to AU coding description of the database, record the expression state of every key frame in each image sequence, then extract the expression parameters through advanced Candide model.(3) Orignizing and designing the semantic system of face expression simulation, constuct the scripting engine system. By making a survey and research of the relation among face subtle motion, region motion and expression motion, we consturct the three-layer semantic system of face expression simulation, then symbolize and parameterize the semantic description. Finally we complete the scripting engine system for face expression simulation underlying above steps.The innovation of the paper is as followed:(1) Propose the novel expression simulation method that combining the MPEG-4 face standard which based on geometric model and face expression code system base on muscle model. It can not only simplify the definition and extraction process of expression paremeters, but also make the expressions generated more realistic.(2) Establish a parameterized face motion model. Parameterized model bulids up the links between high-level semantic description and low-level grid control, and it is convenient for transmission, decomposition and recurrence of face expression because its few grid datas.Base on all of aboved researches, appling the minds of software engineering, finally we implement a face expression simulation system, performs well with few face grid datas.