A Study Of Face Recognition Methods Based On Wavelet Analysis And Support Vector Machine

Automated face recognition is one of the most active and challenging tasks for computervision and pattern recognition, and it is also one of the most important branches of biometrics.Automated face recognition is not only widely applied in a variety of personal identificationsystems such as national security, public security, finance and security facilities, but also canbe used in the fields of visual communication and human-computer interface.Feature extraction and classfier design are two important parts of face recognition.expression, illumination, and pose have bad effect on face recognition, and face recognition isa small sample problem.Considering that the low-frequency sub-graph after the waveletdecomposition has relative insensitivity and the high performance of support vector machinein tackling small sample size,high dimension and its good generalization, this paper focuseson the application of wavelet transform on face feature extraction and the application ofsupport vector machine on face recognition. Based on the classic method, some innovationand improvement are made, and the principal works are listed as follows:(1)Firstly, the basic princilals of the wavelet transform and support vector machine andthe application to face recognition are introduced. Secondly, this passage describes theprincipal component analysis and linear discriminant analysis, and then uses the intergrationof the PCA and LDA algorithm to reduce the dimension and extract the feature of the faceimages. The experiment result on the ORL and YALE face database shows that the method iseffective.(2)A method of face recognition based on two-dimsional weight discrete waveletanalysis and SVM is proposed in this paper.Firstly, the face images are decomposition by thetwo-dimsional DWT, and the low-frequency sub-graph on the second layer is selected only.Secondly, the intergration of the PCA and LDA algorithm is used to reduce the dimension andextract the feature. Finally, the SVM classifier is used for classification. Simulation resultsshow that wavelet decomposition and SVM parameters have effect on the recognition rate.Compared with some other face recognition methods, the experiment result shows the methodachieves a higher recognition rate.(3)In order to make full use of the high-frequency sub-graph after the wavelet decomposition and optimize the SVM parameters, A method of face recognition based onweight discrete wavelet analysis and PSO-SVM is proposed in this paper. Firstly, face imagesare decomposition by the two-dimsional DWT, and then the weightd sum of the waveletcoefficients of four sub-band images on the second layer is selectd as the new waveletcoefficient. Secondly, the intergration of the PCA and LDA algorithm is used to reduce thedimension and extract the feature. Finally, the SVM classifier is used for classification. Theexperiment result shows that the method is effective. Compared with some other facerecognition methods, experiment result shows that the method achieves a higher recognitionrate.