An Anatomically Accurate Biomechanical Model of the Human Face for Simulating Facial Expressions

Abstract

Facial animations are often generated using geometric techniques that disregard the underlying mechanics of facial structures. This thesis presents an anatomically accurate biomechanical model of the face for simulating facial expressions. The computational model uses highly accurate geometric descriptions of skin, facial muscles and the skull surface based on segmented data derived from the Visible Human cryosections published by the National US Library of Medicine. Experimental observations made using an ultrasound imaging system revealed that the superficial soft tissue layer of the face deforms as a single entity. The information acquired led to the creation of a single superficial layer of tissue continuum. Kinematical relations of the computational model were formulated using the theory of finite elasticity to predict large deformations of facial soft tissue structures. The equations governing mechanics of deformable bodies, the static Cauchy equation, and other auxiliary equations were numerically solved using the finite element method. The use of Hermite family elements in the finite element model ensures equilibrium conditions are naturally satisfied, thus more accurate solutions can be obtained. Accurate fibre orientations of facial muscles were derived from reconstructed 3D muscle models and embedded in the continuum. A novel grid based method was employed to determine the activation field of the continuum for simulating contraction of facial muscles. A frictionless contact constraint was imposed via a penalty method to prevent penetrations of facial soft tissues into the skull. An efficient customisation framework was also developed to streamline the processes for generating subject-specific models of facial structures. Furthermore, an investigation was undertaken to assess the suitability of using a coupled membrane-solid model for mechanical analyses of skin and subcutaneous tissues. Simulation results from the coupled model revealed that the assumptions made in membrane mechanics theory lead to significant errors in predicting tissue compressions. Facial expressions were successfully generated for a number of individuals using the optimal activation values inferred from experimental data followed by a validation study to assess the predictability of the model. The detailed biomechanical computational model presented in this thesis has a multitude of potential applications, ranging from facial animation in the entertainment industry to being a clinical tool in facial reconstructive procedures.