For the computation of the physics of the facial skin tissue, we use a mass-spring-damper (MSD) system which has shown to bring the best compromise between accuracy and speed for the deformable objects such as tissue. In generating facial expressions, points on the facial skin are moved in order to obtain the modified face. The energy of the MSD skin can be changed by both the muscle contraction and the jaw action. When facial muscles contract or the jaw activates, the facial skin points that are in the influence area of the muscles or jaw are displaced to their new positions. As a result, the facial skin points not directly influenced by the muscles or jaw are in an unstable state, and unbalanced elastic forces propagate through the MSD system to establish a new equilibrium state. Lagrangian mechanics governs the dynamics, dictating the deformation of facial skin in response to the contraction of the muscle. The relaxation is calculated by solving the governing differential equations numerically. When solving the system of governing differential equations with the numerical method, the number of triangular patches has a direct impact on the computing time and memory consumption. This implies that even with more powerful computers, only facial models with several hundreds of vertices can be dynamically simulated at a rate compatible with real-time graphic animation.Research on physical facial animation has generally relied on explicit numerical integration (such as Euler’s method or Runge-Kutta methods) to advance the simulation. Unfortunately, such methods may lead to slow simulation since very small time steps are required to ensure stability. Moreover, the existing simulation techniques only deal with low-resolution models. Our goal for the numerical simulation module is to develop efficient data structures and algorithms that can process high-resolution facial models. We propose an adaptive simulation algorithm (ASA) by categorizing the skin nodes into different sets based on their dynamic characteristics. The algorithm uses either a semi-implicit integration scheme or a quasi-static solver for the numerical simulation by taking advantage of the facts that facial deformations are local and facial soft tissues are well damped. By propagating forces in an ordered fashion through the facial skin mesh, the governing equation is adapted locally in terms of approximation quality and the computational load is concentrated on the facial regions that undergo significant deformations. ASA runs in real-time and has successfully simulated
realistic facial expressions.
Synthesized typical expressions compared with the actual ones: (a) neutral face; (b) happiness; (c) anger; (d)
surprise; (e) sadness and (f) disgust.
Deformation of the face in simulating different expressions.
Papers:
Yu Zhang, Edmond C. Prakash and Eric Sung. "Face alive". Journal of Visual Languages and Computing, 15(2): 125-160, 2004.
Yu Zhang, Edmond C. Prakash and Eric Sung. "Efficient modeling of an anatomy-based face and fast 3D facial expression synthesis". Computer Graphics Forum, 22(2): 159-169, June 2003.
Yu Zhang, Edmond C. Prakash and Eric Sung. "Real-time facial expression animation on an individualized face using adaptive simulation algorithm". Proc. GraphiCon'2002, pp. 62-69, Nizhny Novgorod, Russia, Sept. 2002.
Yu Zhang, Edmond C. Prakash and Eric Sung. "Hierarchical face modeling and fast 3D facial expression synthesis". Proc. XV Brazilian Symposium on Computer Graphics and Image Processing (SIBGRAPI2002), Best Paper Award, IEEE Computer Society Press, pp. 357-364, Fortaleza-CE, Brazil, Oct. 2002.
Yu Zhang, Edmond C. Prakash and Eric Sung. "Real-time physically-based facial expression animation using mass-spring system". Proc. Computer Graphics International 2001 (CGI2001), IEEE Computer Society Press, Hong Kong, China, pp. 347-350, July 2001.
Copyright 2005-2013, Yu
Zhang. This material may not be published, modified or otherwise
redistributed in whole or part without prior approval.
