Making Faces with Action Unit Morph Targets: 3-Dimensional Parameterized Models of Facial Expression

摘要

The purpose of this study is to create and validate parameterized models for generated life-like facial expressions in a mesh-based animated character. In this study two experimental tasks are used to validate these stimuli. The first task is a simple ratings task where participants rate computer-generated faces and photographs on perceived emotional intensity. In the second task, the faces were rated according to how accurately they convey the intended emotion. Ratings of emotional intensity and accuracy were overall lower for the female animated model. For the male animated model, overall ratings of intensity and accuracy were not significantly different. More confusions between emotions were evident for the synthetic stimuli than for the photographs.