Veering around the Uncanny Valley: Revealing the underlying structure of facial expressions

Summary form only given: When designing both realistic and stylized virtual characters, it is critical to avoid the Uncanny Valley so that visual communication can be consistent, compelling and believable. The human ability to recognize and interpret facial expression as emotion is one of our most sophisticated and innate perceptual skills. This skill is so critical to our life as social creatures that there seem to be multiple neural paths for analyzing faces, a “fine mesh” system for analyzing realistic faces, and a “coarse mesh” system for interpreting human faces seen under difficult viewing conditions, or stylized faces. The fine mesh system relies primarily on analyzing tiny deflections from precise facial norms, while the coarse mesh seems to use pattern recognition rather than position mapping. Certain patterns seem to “trigger” recognition that a certain expression is present.