基于小波變換和支持向量機的人臉識別方法研究.pdf

1、17815加委糾犬學(xué)碩士學(xué)位論文(2005屈)基于小波變換和支持向量機的人臉識別方法研宄Research011faccrecognitionbased。nwavelettrailsforraandsuppogvectormachines研咒生姓名墨堡指導(dǎo)教師姓名—————盞豎直—鱟燕—一專、世名稱———』=焦量焦盤系選——研究方向』|_熊JL塑J置——一論立提變同期—叢生』已L盟————一基于小波變換和支持向量機的人臉識別方法研究Abs

2、tractWiththefastdevelopmentofthetechnologyonthefacerecognitionespeciallyundertherapidspurofcomputer，computer—mdedfacerecognitionhasbecomethelatestoreamongallthefacerecognitiontechnologiesAsanon—rigidobjecthumanfaceshavet

3、hecharacteristicsofvastvariations，beingpronetobeaffectedbymanyinfluentialelements，ThispaperproposestheapproachinintegrationofcombiningwavelettransformandsupportvectormachinesWavelettransforlTlcanfullydemo，nstratethechara

4、ctersofthetargetproblemanditenjoystheoutstandingcapabilityofobtainingcomparativelylesstraitsandinsensitivetothefacialexpressionsinthefacedetectionSupportvectormachines(SVM)arespeciallydevisedtosolvethesmallsampleproblemB

5、ased—onthestructuralriskminimizationprincipleinthestatisticallearningtheorySVMselectstheoptimalseparatehyperplaneaStheseparatefunctionTheoptimalseparatehyperplaneisthehyperplanethateithercorrectlyseparatesthesamplesetorg

6、etsthebiggestmarginbetweentwoclassesnmStheseparateproblemcanbeformulatedasaquadraticoptimizationproblemsatisfiedsimplerestrictionThequadraticproblemhassingularglobalmaximumpointByintroducingthekernelfunction，thenonlinear

7、separatesamplesareprojectedintoahighdimensionspace(socall“featurespace”)，thusseparateproblemissolvedinthelinearsetaratefeaturespaceSVMexhibitssomeparticularsuperiorityfortheclassificationofsmallsamplesandhasbeentheprefer

8、redclassifierattheinternationallevelThethesisuseswavelettransformtoextractfaceimagefcaturethenclassifiesthembasedonSVMRecognitionresultsoftheexperimentsonORLFaceDatabasedemonstratetheeffectivenessofthesystemKeywords：face