Abstract:
Facial expression recognition has been an active research topic for many years, with Facial Action Coding Systems (FACS) being among the widely used methods. FACS is a well-established scheme in psychology to annotate facial muscle contractions and relaxations, also called Action Units (AUs). Previous work on FACS-based methods focused on frontal or near-frontal head poses. In this work, we propose a method to recognize expressions in side head poses. This method builds one classifier for each possible group of occlusions. Facial expression recognition of a side facial pose is then based on a boosting approach of the different classifiers. The method is first tested with frontal and near-frontal head poses, and the results are shown to be comparable to state of the art work for AU and emotion detection. The method is then tested with a small training set for various orientations and AUs, and shown to be accurate. © Springer-Verlag 2012.