Aesthetic exploration and refinement: A computational framework for expressive character animation.

摘要

Character animation remains a very challenging task despite many years of research on improved tools and algorithms. Production-quality animation is extremely slow to produce and only a small number of skilled animators are capable of creating it. This thesis explores methods to ease the creation of character animation, with a particular focus on expressive aspects of movement that convey personality and mood. The goal of the work is to identify which aspects of movement can be productively given computational representation and how this can best be done; its intent is not to develop a particular user interface. To reduce the scope of the problem, a restricted set of standing movements---including gestures, posture changes and weight shifts---is used as a test bed. An extensive review of the performing arts literature was conducted to understand which aspects of movement are most salient to expression. This study produced a list of key expressive aspects of movement. Computational representations for many of these movement properties have been developed, giving them a precise semantics. In addition, a software framework has been built that allows the various movement properties to be combined in a consistent and predictable manner by first mapping them to a low-level motion specification that is then simulated either kinematically or dynamically. A key feature of the system is a range of input modalities, including: character sketches that allow global changes to a motion sequence, high-level properties that adjust many aspects of a particular action, and low-level properties that precisely control particular attributes. These modalities support both rapid exploration of the aesthetic space and detailed refinement. Directly representing these resources in an easy to understand manner allows them to be customized and extended to meet the needs of a particular animator or a particular character. An implementation and various animations have been generated to illustrate the effectiveness of the approach. This work represents a step towards creating a general language for movement that can act as a bridge across artistic and technical communities to resolve ambiguities in the discussion of motion.