The Facial Action Coding System (FACS) [Ekman & Friesen 1978] is a system designed for empirical coding of human facial expressions. The FACS model is based on a simplification of the muscle actions involved in producing human facial expression, where muscles are grouped together into what the authors call Action Units. Waite [1989] modeled a human face based on a control structure that incorporates several of the action units described in ฃ Ekman & Friesen [1978]. In her system, the action units are represented by collections of data points which are covered by a single rendered surface that mimics human skin. The approach taken does not automatically solve how to draw the eyes, control gaze, or add other decorative features (such as ears or hair) to the rendered face. Because the system relies on a model of muscles and bone structure, it is computationally intensive. More recently, Takeuchi & Nagao [1993] describe a system that tries to model a real face in three dimensions based on a similar approach, and Essa [1995, Essa et al. 1994] describes a computational extension to FACS.