基于特征融合的笑臉識(shí)別算法研究.pdf

1、西安電子科技大學(xué)碩士學(xué)位論文基于特征融合的笑臉識(shí)別算法研究姓名：嚴(yán)新平申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：通信與信息系統(tǒng)指導(dǎo)教師：盧朝陽(yáng)201201AbstractSmilerecognitionhasrecentlyattractedsignificantattentionduetotheincreasingaccessibilityofinexpensivedigitalproductsandthedevelopmentofpatternrec

2、ognitiontechnologyMoreoversmileisregardedasallimportantexpressionandtheimprovementofsmilerecognitionCanpromotethedevelopmentoftheexpressionrecognitionAsaresult，improvingtheperformanceofsmilerecognitionhasbecomeavalualble

3、taskThemaincontentsofthisdissertationareasfollows：1ThesmilerecognitionbasedonthefeatureoffaceisstudiedTheLBPasthedescriptorforthefaceisselectedGiventhattheinherentcharacteristicsofsmileexpressionthefaceblockweighedLBPfea

4、turesforsmilerecognitionareusedcombinedwiththepreciseofthefacelocalizationTheexperimentalresultsshowthatthismethodissuitableforthestrongexpressions2ThesmilerecognitionbasedonthefeatureofmouthisstudiedThemouthlocalization

5、andthedescriptionofthemouthfeaturearediscussedindetail：theapproximatemouthregioncallbeobtainedbasedonthegeometryfeaturesofface，andthentheaccuratemouthregionCanbeobtainedbyOstumethodcombinedwimthehistogramspecification；Th

6、eHOGasthedescriptorforthemouthisselectedTheexperimentalresultsshowthatthismethodperformswellcomparedtothefirstmethodforthesmileexpression3ThesmilerecognitionbasedonthefeaturefusionisstudiedTheresearchaboutthefusionalgori

7、thmsbasedonseriesconnectioncanonicalcorrelationanalysis(CCA)，andthediscriminativeCCA(DCCA)arediscussedindetailTheexperimentalresultsdemonstratethattheeffectivenessandsuperiorityofthefusionbasedonDCCA4Thethreefeaturerepre

8、sentationsareusedtoexpressionrecognitionGiventhatthedifferencebetweentheexpressionrecognitionandthesmilerecognition，thefeatureselectionandthemulticlassificationofthesupportvectormachine(SVM)areanalyzedchieflyAndtheperfor