Research On Face Recognition Based On Subspace

Face recognition has on many occasions the important role that traditionalauthentication methods such as using some sort of identification number, but it isdifficult to prevent the occurrence of counterfeiting. As the human visual characteristics,such as face, gesture and so is relatively stable and varied, so as with the identificationof these features is feasible. Samples of face recognition technology often encounterhigh dimension, a large number of categories, the training sample size, as well as pose,illumination, facial expressions of other issues.In this paper, it’s a problem that ICA-based subspace and linear regression methodof combining. Face and positive attitude of people using face feature vectors to achievethe transformation matrix between the attitude profile face feature vector transformation,then transform the feature vectors based on the synthesis of multi-frontal face andgesture recognition. In this paper, polynomial transform and sine transform method toincrease the virtual samples increased by increasing the recognition rate of trainingsamples. Increasing virtual samples, within class scatter matrix to solve the problem ofzero, making the criterion based on the various methods of linear-on-one trainingsample face recognition problem can also be used. The presented computer simulationsand numerical examples also show the efficiency of these methods.This paper introduces the basic principle of principal component analysis andapplication in face recognition. The paper also discusses the advantages and disadvan-tages of PCA and gives2dimensional PCA in face identification in the experimentalsteps. For the face of the sample data in the interference condition, this paper givens themultiple posture face identification method for face identification approaches.Numerical simulation and experiment also shows that these methods are effective.