Research On Subspace Analysis Based Face Recognition Algorithm

The recognition of human faces is an active subject in the pattern recognition field in the past few years, and it has a wild range of potential applications. But until now, research results that we have got are far away from satisfactorily solved.Firstly, the background, content, significance of the human faces recognition field are summarized. The subject methods of the human faces recognition are introduced one by one. Secondly, two methods of human faces recognition based on linear subspace analysis have been given a detailed description. They are Principle Component Analysis and Principle Component Analysis based on Fisher criterion.A new method of Principle Component Analysis based on Singular Value Decomposition (SVD) and Fisher criterion is presented. We use singular value vector to describe images, then singular value vector is considered as feature vector for recognition. Because singular value decomposition has good properties in mathematics, the method has an advantage in recognition rate than serialization of image.Experiments are performed on human face library which depend on Principle Component Analysis based on Singular Value Decomposition (SVD) and Fisher criterion. The results have been analyzed and contrasted.