Mithraratne, KHunter, PWu, Tim2014-01-232013https://hdl.handle.net/2292/21457This thesis presents a computational framework for modelling the biomechanics of human facial expressions. The expressive motions are formed by intricate interactions of muscles and other soft tissues present in the face. In order to elucidate this complex mechanism, an anatomically realistic finite element (FE) model of the face was created from a set of highly detailed magnetic resonance imaging (MRI) data. A volume-preserving free-form deformation algorithm with 3-D surface data acquired from a structured-light scanner was used to transform the MRI-derived model from a supine position to an upright realis-tic setting. The biomechanics of the detailed facial model was characterised by the theories govern-ing finite deformation elasticity and nonlinear contact mechanics. The elastic difference between the muscle and adipose tissues was accounted for by treating the tissue continu-um as a heterogeneous medium and its mechanical behaviour was represented using a two-parameter Mooney-Rivlin constitutive model. In order to make the constitutive equa-tions compatible to a mixed (displacement-pressure) FE formulation for incompressible bodies, a modified version of the Mooney-Rivlin model was implemented. Soft tissue de-formations were driven by the contraction of the underlying network of 3-D muscles with accurate depiction of their fibre orientations and mechanics. Frictionless unilateral and sliding contact constraints were imposed to model the interaction between the superficial soft tissue continuum and deep structures. The resulting system of nonlinear equations was solved numerically using the Netwon-Raphson method with a trust-region implemen-tation. Visually realistic facial expressions and emotional gestures were simulated using the biomechanical model with suitable muscle activation levels. In order to achieve real-time performance for the expression prediction, a workflow based on statistical surrogate models was presented. Using the multivariate partial least squares regression model, it was demonstrated that the facial biomechanics can be emu-lated efficiently in real-time with reliable solutions. It was also demonstrated how these surrogate models can be implemented for the inverse problem, where the muscle activa-tion patterns can be estimated from the experimentally determined facial expressions. Incorporating features at both the organ and tissue levels provided a framework for ac-curate and reliable prediction of facial biomechanics, as demonstrated in this thesis. Moreover, further acceleration can be achieved by emulating biomechanical behaviour using statistically-based surrogate models.Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated. Previously published items are made available in accordance with the copyright policy of the publisher.https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htmhttps://creativecommons.org/licenses/by-nc-sa/3.0/nz/ThesisCopyright: The Authorhttp://purl.org/eprint/accessRights/OpenAccessQ112904205