FACEFACTS: Modeling Natural Facial Expressions

Instead of simply assuming a representational basis developed for humans examining static imagery (e.g., FACS) our approach to expression modeling is to find a set of basis unit for expressions that can be extracted from natural data with high accuracy and then to use these basis expressions to build a representation of higher-level, complex expressions. In this section we describe all the elements of our proposed framework for analysis and modeling of facial expressions. We start with the design of our data collection environment. Then, we elaborate on the head-tracking, feature extraction and modeling methods that allow us to study and analyze naturally expressive data.