预览加载中,请您耐心等待几秒...
1/8
2/8
3/8
4/8
5/8
6/8
7/8
8/8

在线预览结束,喜欢就下载吧,查找使用更方便

如果您无法下载资料,请参考说明:

1、部分资料下载需要金币,请确保您的账户上有足够的金币

2、已购买过的文档,再次下载不重复扣费

3、资料包下载后请先用软件解压,在使用对应软件打开

SparseRepresentationforFaceRecognitionbasedonDiscriminativeLow-Rank DictionaryLearning LongMa,ChunhengWang,BaihuaXiao,WenZhou StateKeyLaboratoryofManagementandControlforComplexSystems InstituteofAutomationChineseAcademyofSciences 95ZhongguancunEastRoad,100190,BEIJING,CHINA {long.ma,chunheng.wang,baihua.xiao,wen.zhou}@ia.ac.cn Abstractmizationproblem: minxs.t.y=Dx(1) Inthispaper,weproposeadiscriminativelow-rankx1 dictionarylearningalgorithmforsparserepresentation. whereDisanover-completedictionary,xisthesparseco- Sparserepresentationseeksthesparsestcoefficientstorep- efficientvector,andyisthetestsignal.Toimprovetheper- resentthetestsignalaslinearcombinationofthebasesin formanceofsparserepresentation,Yang[30]proposedro- anover-completedictionary.Motivatedbylow-rankma- bustsparsecodingtomodelthesparsecodingasasparsity- trixrecoveryandcompletion,assumethatthedatafromthe constrainedrobustregressionproblem;Liu[21]constrained samepatternarelinearlycorrelated,ifwestackthesedata thesparsecoefficientstobenonnegative;Huang[14]ex- pointsascolumnvectorsofadictionary,thenthedictio- ploitedtheclusteringtendsinnonzerocoefficients.These naryshouldbeapproximatelylow-rank.Anobjectivefunc- algorithmsusedtheoff-the-shelfbasesasthedictionary. tionwithsparsecoefficients,classdiscriminationandrank Learningthedictionaryhasbeenprovedtoimprovethesig- minimizationisproposedandoptimizedduringdictionary nalreconstructiondramatically[8].Severalalgorithmshave learning.Wehaveappliedthealgorithmforfacerecogni- beenproposedtooptimizetheatoms.Aharon[1]general- tion.Numerousexperimentswithimprovedperformances izedthek-meansclusteringprocessandproposedK-SVD overpreviousdictionarylearningmethodsvalidatetheef- algorithm,thealgorithmiterativelyupdatedthesparsecod- fectivenessoftheproposedalgorithm. ingofthesamplesbasedonthecurrentdictionaryandthen optimizedthedictionaryatomstobetterfitthedata.Mairal [23]proposedanenergyformulationwithbothsparsere- constructionandclassdiscriminativecomponents.Anon- 1.Introd