Analyse de visages et d’expressions faciales par modèle actif d’apparence
In this paper, methods are proposed for facial feature detection (eyes, brows, nose, mouth, chin) and for facial expression recognition. The methods are based on modified versions of the standard Active Appearance Model proposed by Cootes et al. to control both the shape and the texture of a given face. The detection algorithm makes use of an active appearance model computed on hierarchical Gabor descriptions a set of training faces. In a second part, two expression models are proposed, based on the standard AAM, and used to recognize and then to cancel or modify the facial expression of a given unknown face.