Face Recognition Based On Manifold Learning And Classification Technology

Face recognition is a very attractive technology which has been widely used in many fields in recent years.Face recognition works in the condition that the facial characteristics information has been already given.Firstly,face recognition technology do the work of imitation and the sample collection to build a face database.In recent years,with the development of science and technology,the face recognition technology has been widely applied in all aspects and becomes a well concerned field in today's studies.As for the research of this technology,several methods have revealed progressive results,such as the traditional classification algorithms support vector machine(SVM)and extreme learning machine(ELM).However,face data collected is usually of high dimension as well as data distribution is also imbalanced in the process of face recognition.Therefore,how to solve the problem of unbalanced high dimensional data in face recognition will become a hot spot of research.Based on SVM and ELM,by exploiting manifold theory and differential homeomorphism theory,two new face recognition algorithms are proposed in this study:Manifold SVM algorithm.During the face recognition process,SVM can cause the fault points increasly due to high dimensional data,as well as reduce the face recognition rate.So,a face recognition algorithm based on fuzzy clustering locally linear embedding and SVM is proposed.The fuzzy clustering of the locally linear embedding algorithm is introduced on the basic of SVM,which can better constraint reconstruction error and keep the inner structure of data.From the experimental results,it can be seen that the new algorithm can better deal with high dimensional data,so as to improve the recognition effect.Manifold ELM algorithm.Based on ELM,we proposed a method to construct activation function by using the theory of differential homeomorphism,which can conserve the nature of the data.In the structural formula of NPE,the information of between-class and within-class is adopted to maintain the data structure.This algorithm also can eliminate the redundant data of samples.The experimental results show that new proposed algorithm have a higher capacity of classification.