Articulated Figure Animation

In the early seventies the notion that computers could be programmed to create images worthy of being called "realistic" was a dream brought closer to reality by diligent experimentation with models and with light and display algorithms. Within a decade great progress was demonstrated by the almost commonplace production of high-quality images. But while the most complex real or fantastical objects had been visualized and even animated with computer graphics, certain common objects-people and animals-seemingly eluded the best efforts to capture their form, fluidity, and expression. The belief that computer graphics should be able to animate synthetic people or animals was almost a romantic vision, comparable to the quest for image realism of the early graphics pioneers. It seemed easy enough. After all, weren't people and animals just combinations of rotary joints which could be controlled through the right choices of joint angles, key positions, and interpolatory schemes? While the animation repertoire is full of such experiments, they almost universally fail to convey much convincing evidence that the computational model is close to achieving synthetic natural animacy and spontaneity. When motion information is measured directly off human subjects, the result is natural motion, but with little theory of how such motion can be synthesized. When figure animation is successful, it is most often because a visionary and sensitive animator has overcome the weakness of the tools to create a successful character or mood. Five years ago I was Guest Editor for a special issue of Computer Graphics and Applications (November 1982) on "Modeling the Human Body for Animation." Though it appeared to emphasize modeling, in fact significant emphasis was placed on the animation control issues, including seminal descriptions of models based on simulation, finite state machines, parametric interpolation, and goal-directed kinematics. In this issue I revisit this topic with a different emphasis. Now the control issues clearly dominate the articles, as modeling is understandably perceived to be more a problem of data manipulation (which can be tackled through various surface-modeling or solid-modeling schemes), while motion is the essence of the communication. One need only observe excellent conventional (noncomputer) animation to realize that motion fidelity is more important than shape fidelity in conveying mood, expression, and character. In seeking better motion control, computer animation researchers have established a number of methods apparently essential, or at least effective, in creating certain classes of motion. Those methods are presented in the articles of this special issue.