A Blendshape Model that Incorporates Physical Interaction

How to read this

Category

Method: a physics-augmented blendshape facial model

Contributions

• Augments linear blendshape interpolation with a physically based mass-spring system to fix degradation under large rotations and the inability to handle collisions
• Builds a mass-spring system per blendshape target, interpolates spring rest lengths, and solves for the equilibrium shape (minimizing local area and volume distortion) for plausible deformation across large rotations
• Enables physical interaction with other scene elements via collision detection and handling, with no precomputation, skeleton, or manual intervention

Context

Builds on blendshape facial animation and direct-manipulation blendshapes (Lewis and Anjyo 2010), adding a physical equilibrium layer to address linear-blending artifacts.Builds on: Direct Manipulation Blendshapes

Correctness

Plausible deformation across large rotations and contact is the claim; results are physically based but use a mass-spring (not full FEM) model so realism is approximate, and a reader should note quality depends on the per-target spring system and on collision detection robustness.

Clarity

Clearly motivated by two known blendshape failure modes; a first pass conveys the rest-length-interpolation-plus-equilibrium idea, with a second pass for the mass-spring setup and solve.

How to read it

First pass for why linear blends fail and how the equilibrium reformulation helps; second pass on the mass-spring construction and collision handling if extending or implementing it.