Font Size: a A A

# Research On Feature Extraction Based On Principal Component Analysis

Posted on:2017-08-07Degree:MasterType:Thesis
Country:ChinaCandidate:X H ZhaoFull Text:PDF
GTID:2348330542450151Subject:Computational Mathematics
Abstract/Summary:PDF Full Text Request
As a common method used to extract data feature,principal component analysis(PCA)has been widely applied to pattern recognition,image processing,data mining and other fields.This paper studied the data feature extraction based on principal component analysis method.The work has been done as follows:Firstly,it is a difficult problem to select the number of eigenvector in data feature extraction based on principal component analysis method.The popular method is selection of threshold by experience,then selection of feature vector based on the cumulative contribution rate of eigenvalue.In order to solve the selection problem of feature vector and extract the data feature more effectively,the recursive two dimensional principal component analysis(R2DPCA)is proposed on the basis of the two dimensional principal component analysis(2DPCA).The main aim is to improve the SNR of data feature by recursive algorithm.The experimental results on the public face database show that the proposed method is more effective than the method of two-dimensional principal component analysis.Secondly,kernel principal component analysis(KPCA)is one of the mature kernel methods,which can extract nonlinearity of the data by the kernel technique.However,selection of parameters of kernel function is very importance.The popular methods are trial and error method and leave-one-out often used for selection of parameters of kernel function.Those methods are time-consuming,and the selected parameters of kernel function not necessarily meet the actual needs.This paper take radial basis function as the function of KPCA,and the method of nuclear integrating principal component analysis algorithm is proposed.Experiments on the standard face database verify that the algorithm is more effective to extract data features.Thirdly,calculation of eigenvector is one of the main factors which restrict the algorithm of KPCA.Under the condition of large sample size,the method of KPCA to obtain the eigenvalue and eigenvector is particularly complicated.Usually matrix decomposition method can be used to get the eigenvalue and eigenvector.So nystr?m method is used to approximate the eigenvector.Nystr?m method is seriously restricted by the extracted samples.In order to overcome the randomness of Nystr?m method,this paper puts forward the method of Nystr?m integrated KPCA method.Experiment show that the method of Nystr?m integrated KPCA method is better than the Nystr?m method used in KPCA for data feature extraction.
Keywords/Search Tags:principal component analysis, recursive two-dimensional principal component analysis, parameters of kernel function, kernel principal component analysis, kpca integration algorithm, nystr?m integrated kernel pca algorithm
PDF Full Text Request
Related items
 1 A Weighted Kernel PCA And The Related Parameters Choice 2 Recognition Method Of Metal Fracture Images Based On Empirical Ridgelet And Principal Component Analysis 3 Recognition Methods Of Metal Fracture Surface Images Based On Grouplet Transform And Kernel Methods 4 An Application Research On Face Recognition Based On KPCA Feature Extract And SVM 5 Research On Feature Transformation Based On Kernel Principal Component Analysis 6 Construction Method Of Principal Component Networks And Its Application 7 Research Of Image Denoising Methods Based On Kernel Principal Component Analysis 8 Fault Detection And Diagnosis Based On Kernel Principal Component Analysis 9 Welding Defects Recognition Algorithm Reasearching Based On Principal Component Analysis 10 Novelty Detection Based On Robust Kernel Principal Component Analysis And Kernel Entropy Component Analysis