Study On Statistical Model In Facial Expression Recognition And Its Applications

Recently, face recognition is a hot topic on pattern recognition. Many methods have been developed in the literature. General speaking, face recognition contains two steps: face localization and classification. Exact classification is based on extract localization, both of them are important issues in face recognition. But the feature abstract of face is the most difficult and popular problem in these area. The feature abstract of face is the core technology of human face recognition. It consists of computer technology, image process technology and pattern recognition technology and so on. This biologic technology of recognition is developing quickly and widely. In this dissertation, nearest neighbor classification(NNC) and k-nearest neighbor(KNN) are integrated with active shape model(ASM) and active appearance model(AAM). The main contributions are summarized as follow:(1)Classical ASM,AAM,combined model and improved methods are studied. A Unified Model (AAM+ASM) is an efficient method for extracting human facial features. It includes human facial dynamic appearance models and the fitting algorithm. This paper proposes an improved in the algorithm. On the basis of the unified model,This paper make the best use of the relation of points for develop a new method constructed local appearance model. The improved algorithm adds the errors of local appearance model, and it made the deformation more remarkable after every iterate in the fitting process, so the fitting speed been improved. The improved algorithm make the best use of the relation of points,and get better texture information about the points and it made the fitting result closer to the input image,so the precision of feature extraction has been improved。The experiments indicate the effectiveness of the improved algorithm, the improved fitting algorithm improve the precision of the location of the feature points and the fitting speed.(2)A method combined combined model and NNC are proposed and used in face classification. Locate coordinate acquiring by combined model is used as the input of the classification method。...