基于面部解剖结构动力学模型与多模态时空数据耦合的人脸仿真

Facial motion simulation is not only a grand challenge but also a hot research issue in the field of computer graphics. However, generally, existing methods have problems such as one-sided considered construction for geometric model, single modal data for data driving and dynamic model in need of improvement. Aiming to solve the above problems, this project is to start form the research on adaptive realistic Bone-Muscle-Skin (BMS) structure, consider multi-modal spatio-temporal data coupling as a benchmark, explore the nonlinear dynamic model of BMS structure, utilize a nonlinear deformation model to improve the existing mass-spring model (MSM) and achieves an improved MSM (IMSM) with constant volume constraint, and establish a unified framework of multi-modal spatio-temporal data coupling dynamic model for BMS . Through the research in the project, it’s expected to get a normative facial model based on anatomical structure, provide a multi-modal spatio-temporal data coupling mechanism and build a unified theory for facial dynamics from physics viewpoint. The successful completion of this project can promote not only the in-depth development of multiple related applications, but also the further research of the facial expression simulation and related fields with new ideas and theoretical basis.

目前人脸运动物理仿真是计算机图形学研究领域的重大挑战问题，也是研究热点问题。然而现有方法普遍存在几何模型构建片面、驱动数据模态单一以及动力学模型亟待改进的问题。本课题针对以上问题，以自适应真实感骨骼-肌肉-皮肤（Bone-Muscle-Skin,BMS）结构为研究出发点，以多模态时空数据耦合为基准，探究BMS结构的非线性动力学模型，利用非线性变形模型改进现有的质点-弹簧模型（Mass-Spring Model，MSM）建立体积不变约束的改进MSM（Improved MSM，IMSM），建立统一的多模态时空数据耦合的BMS动力学模型框架；通过本项目的研究，有望得获得准确规范的基于解剖结构的面部模型，给出多模态时空数据耦合机制以及构建统一的面部物理动力学理论。本项目的顺利完成能促进相关应用领域的深入发展，并为推动面部表情仿真及相关研究的深入提供新的思路和理论基础。

目前人脸运动物理仿真是计算机图形学研究领域的重大挑战问题，也是研究热点问题。然而现有方法普遍存在几何模型构建片面、驱动数据模态单一以及动力学模型亟待改进的问题。本课题遵循项目技术路线和研究方案，研究了真实感骨骼-肌肉-皮肤（Bone-Muscle-Skin,BMS）结构构造方法，提出了使用Thin-Shell结构肌肉模型提升BMS仿真效率，基于体积不变约束给出了运动仿真模型，探索了统一的多模态（包括RBF分区数据、运动捕捉数据与FACS参数数据等）时空数据耦合框架。通过项目研究，已写成多篇学术论文，还包括《计算机图形学》双语教学研究论文，形成发明专利两项。项目成果满足面部医疗、面部可视化、模式识别、影视游戏等多领域应用的深层次需求，并为面部仿真和相关研究领域提供新的思路和理论依据，具有的推广价值。