资源描述:
《Evaluation of a Case-based Facial Action Units Recognition Approach》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、EvaluationofaCase-basedFacialActionUnitsRecognitionApproachS.F.Wang,J.Xue,X.F.WangDept.ofComputerScienceUniv.ofSci.andTech.ofChina(USTC)Hefei,Anhui,230027,P.R.Chinasfwang,xfwang@ustc.edu.cn,xuejia@mail.ustc.edu.cnAbstract—Inthispaper,weevaluatetheperform
2、anceofaexpressions.However,achievingAUrecognitionbyacom-case-basedautomaticfacialactionunitsrecognitionapproachputerremainsdifficult.usinginteractivegeneticalgorithm(IGA).First,thecase-basedfacialactionunitsrecognitionapproachisintroduced.ThisFewapproache
3、shavebeenreportedforautomaticrecog-methodretrievesthemostsimilarcaseimagefromcasedatabasenitionofAUsinimagesoffaces[4][5][6][7][8][9].ThemainusingIGAandreusestheactionunitsofthematchedcaseimageframeofthesepreviousworksisfirstlytoextracttheparam-tothetestf
4、aceimage.Second,toevaluatetheeffectivenessofetersdescribingthefacialfeatures,andthentorecognizedourapproach,comparisonexperimentswitheigenfacemethodAUsusingmachinelearningtechniques.Inordertomakeonsimpletestimagesaredone.Theexperimentalresultsshowthat,fo
5、rourmethod,theaveragerecognitionrateisabouttheproblemoffacialfeaturedetectionmoretractable,most77.5%onsingleAUsandaveragesimilarityrateis82.8%onAUofthepreviousworkreliedondata-setofposedexpressionscombinations,whicharebothhigherthanthoseoftheeigenfacecol
6、lectedundercontrolledimagingconditions,e.g.,withsub-method.Third,experimentsofthecase-basedautomaticfacialjectsdeliberatelyfacingtothecamera,withnofacialhairactionunitsrecognitionapproachoncomplextestimagesisorglasses,andwithconstantillumination.Thisison
7、eofpresentedinthispaper.Theresultsprovetherobusticityofourapproach.ArecognitionrateofsingleAUsof82.8%andathemajorimpedimentstoextendthesesystemstorealisticsimilarityrateofAUcombinationsof93.1%areobtained.application.Incontrasttothosepreviousapproaches,th
8、eresearchKeywords—facialactionunits,recognition,IGAreportedhereaddressestheproblemofsemiautomaticAUrecognitionusingIGAdirectlywithoutthestageoffacialI.INTRODUCTIONfeaturedetection,whichissensitivetoimagingconditions.Thecor