A new physical model with multilayer architecture for facial expression animation using dynamic adaptive mesh

Area of Science:

• Computer Graphics
• Biomechanical Modeling
• Animation

Background:

• Existing facial models often struggle with realistic soft tissue deformation and computational efficiency.
• Mass-spring-damper (MSD) models lack the nuanced simulation of visco-elastic properties.
• Accurate anatomical representation is crucial for believable facial animation.

Purpose of the Study:

• To develop a physically-based 3D facial model that achieves high-fidelity expression animation.
• To optimize computational performance for real-time or near-real-time applications.
• To incorporate anatomical knowledge for more realistic skin and muscle interactions.

Main Methods:

• A multilayer biomechanical structure simulating skin visco-elasticity with nonlinear springs.
• Anatomically-motivated muscle actuators and a skull structure to constrain skin movement.
• A local semi-implicit Ordinary Differential Equation (ODE) solver with adaptive mesh refinement for dynamic simulation.

Main Results:

• The model accurately simulates nonlinear visco-elastic soft tissue behavior and prevents skin model collapse.
• Incorporation of skull structure and muscle models leads to more accurate facial deformations.
• Adaptive refinement significantly reduces computational cost by focusing resources on areas of high deformation.

Conclusions:

• The proposed physically-based 3D facial model offers a significant advancement in realistic and efficient facial expression animation.
• The integration of biomechanical principles and adaptive computation provides a robust framework for complex facial dynamics.
• This approach enables the production of more pleasing animation results at a reduced computational expense.