A Single Training Sample Face Recognition Based On Modular Weighted (2D)~2PCA

The person face is the most common image in human vision;face recognition is the most important aspect in human vision.For a long time,how to use computer for automatic face recognition has bean a hot spot and difficulty in imagery processing and pattern recognition.Face recognition with a single sample per person,as the name suggests,is to identify an unknown testing sample under the circumstance that every subject in the training sample set has only a single sample.Specially,in the situation of one training sample,face recognition is more difficult.Based on the research of classical face methods,used a Two Dimensional Principal Component Analysis method union horizontal,row direction,unioned modular and weighting thoughts,this paper has carried on the feature extraction of the training sample.In recognition stage,by comparing the training samples and the characteristics of the unknown sample,determine the identity of unknown samples,complete identification.The main research work of this article is as follows:1.Discussed the purpose and significance of the single sample face recognition. the present research situation at home and abroad.Inductived the method commonly used in face recognition.Analysed characteristic subspace extended method2.Based on the classic principal component analysis of single-sample face recognition,this paper discussed PCA,KPCA and SPCA in detail.Discussed the principle and implementation steps of these algorithms.3.Analysed a method union horizontal,row direction.Made the improvement of the algorithm.Studied a modular weighted(2D)~2pCA.On the one hand, outstanded the influence of difficult eigenvalue correspondence eigenvector for the result of the distinguish through weighting.On the other hand,carried on the modular image,so that more effective local features could be extracted In recognition stage,according to maximum membership degree law,determine the identity of the unknown image. 4.In Windows XP platform,did a large number of experiment in ORL face database.Carried on PCA,KPCA,SPCA and the method used in this paper.Analysed the influence of two parameters in SPCA algorithm.Emphatically analysed the influence of weight and modular for recognition rate.Gived the most superior weight and the best modular mode.Compared the recognition result of the method used in this paper and PCA,KPCA,SPCA and other methods in ORL face database.The result of the method used in this paper was the best.