Automatic Facial Animation Analysis

The primary aim of automated facial animation is to allow us to take a 3D head model of a subject in a single expression (usually neutral) and to generate a range of expressions and facial poses without requiring the intervention of a traditional animator. Given the ability to automatically generate a full range of expressions and head poses for a specific model we are able to use the model as a “talking head” for enhanced communication, within training scenario software, for use in games or movies or for face / expression recognition applications. It is highly desirable to fully automate the animation process since this reduces the need for animators to define key frames whilst vastly reducing the amount of time required to produce realistic virtual characters. Considering video games for a moment, in-game characters must respond dynamically and realistically to actions presented by the player. In a game that allowed highly detailed face models the large number of potential expression permutations may be difficult and time consuming to animate by hand. Furthermore scenarios in virtual worlds where a player may wish their in-game avatar to visually represent their real life appearance would benefit greatly if the system was able to reconstruct a head model and then accurately animate expressions mapping a players emotions to their virtual avatar. Finally, assuming a suitable model and animation structure it would be possible to map animations from one head model to another with little or no work, thus allowing facial animations to be reusable across a large range of head models.