Spotting the Details: The Various Facets of Facial Expressions

Please always quote using this URN: urn:nbn:de:0297-zib-67696
  • 3D Morphable Models (MM) are a popular tool for analysis and synthesis of facial expressions. They represent plausible variations in facial shape and appearance within a low-dimensional parameter space. Fitted to a face scan, the model's parameters compactly encode its expression patterns. This expression code can be used, for instance, as a feature in automatic facial expression recognition. For accurate classification, an MM that can adequately represent the various characteristic facets and variants of each expression is necessary. Currently available MMs are limited in the diversity of expression patterns. We present a novel high-quality Facial Expression Morphable Model built from a large-scale face database as a tool for expression analysis and synthesis. Establishment of accurate dense correspondence, up to finest skin features, enables a detailed statistical analysis of facial expressions. Various characteristic shape patterns are identified for each expression. The results of our analysis give rise to a new facial expression code. We demonstrate the advantages of such a code for the automatic recognition of expressions, and compare the accuracy of our classifier to state-of-the-art.

Download full text files

Export metadata

Additional Services

Share in Twitter Search Google Scholar
Author:Carl Martin Grewe, Gabriel Le Roux, Pilz Sven-Kristofer, Stefan Zachow
Document Type:ZIB-Report
Date of first Publication:2018/03/07
Series (Serial Number):ZIB-Report (18-06)