A Novel Approach For Projection By Optimize Local Area In Face Recognition Applications

With the rapid development of science technology,there are various of identification technology is widely attention by humans. As an important symbol of distinguish human characteristics, face plays a vital role, its attention in recent years increased significantly. Face recognition technology had many characteristics, including high stability, strong intuition, and rapid safety, and used in various fields,so it also became a hot issue in today's research field. As we all know, in the face recognition study, first we should do the extraction of face characteristics, this is the key to solve this problem. While the computer storage are digital image,here we facing an important problem is how to confront high-dimensional data processing.,namely dimension disasters. How to use dimension-reduction methods effect quickly to solve this problem, manifold learning gave us a lot of inspiration. Through the way of manifold learning, we can seek from high-dimensional data to the corresponding low dimensional manifold structure, Thus effectively deal with the problem of face recognition.This paper is mainly introduces the classic dimension-reduction methods, including linear method and nonlinear method, from it we caught the essence of dimension reduction, in traditional dimension-reduction methods, people reduce dimension subjectively, and lost dimension related important information, projection vector as constituting the projection matrix's main ingredient,it determines the projection results. The work of this paper firstly adopts the traditional dimension-reduction methods to get projection vector, using the classical algorithm ADABOOST for projection vector optimization again,we call it part projection vector.and use it for projection.This adopt ADABOOST with local optimization idea reflect the effectiveness of the method.Through the simulation test, the paper in many database had a good experimental results. Meanwhile, the paper also introduces two other kinds of optimization algorithm, genetic algorithm and particle swarm algorithm, through the contrast of paper, we verify the mentioned the feasibility and effectiveness of the approach.