|
Kevin[at]thekrf.com |
Motion Curves : A versatile representation for motion data
Abstract
This thesis presents Motion Curve space: a novel representation scheme for the poses of an articulated skeletal figure. A Motion Curve space is defined by a set of orthogonal basis vectors that have been found by performing a weighted principal component analysis on an example motion clip. An animator can control the properties of the space through the selection of the example clip and the PCA weights. We explore the expressive and computational power of the representation through the creation of several new motion processing and analysis algorithms, which are demonstrated through prototype applications. These prototypes help to establish the workflow for a hypothetical production application. In presenting this work, we hope to expand the size of the animator’s toolbox. By providing a new and usable framework for editing motions, we make it possible to quickly modify existing motion assets and stretch animation budgets.
Some source code (very messy and possibly incomplete - use at your own risk :)