Study On Improved Active Appearance Model For Face Recognition

As a main research field of biometric identification, face recognition has been studied intensively with high utility value. Compared with other biometric technologies, face recognition has much more advantages. But in the unconstrained environment there are still many problems. The results of the recognition can be seriously affected by the inner and outer factors such as expression, attitude, light conditions and background, Active Appearance Model (AAM) is the expansion based on the Active Shape Model (ASM). AAM uses the priori model to match the face in the image. The advantage is the model that is not only contains the shape information but also contains the texture information. This can be used in marking feature points, synthesizing face images, tracking object and so on.In this paper, an improved AAM method for face recognition is studied. AAM is used to synthesize new face images. The shape model of AAM, which express the variation of face attitude, can be removed so that the face images are registered to new synthesized images without the attitude influence. By improving AAM, we reduce some inner and outer effects on recognition process, and the attitude especially can be removed in order to obtain a better result. Experiments which use three schemes, LBP with support vector machine (SVM), LBP with linear discriminant analysis (LDA), and HMAX with SVM, are carried on IMM database respectively for the original images and the synthesized images. The results indicate that the HMAX combined with SVM for AAM synthesized face images gets a better recognition rate.