简介：Inthispaper,anewtechniqueisproposedforrotation,scalingandtranslation(RST)invariantimagewatermarkingbasedonlog-polarmappings(LPM)andphase-onlyfiltering(POF).ThewatermarkisembeddedintheLPMofFouriermagnitudespectrumoftheoriginalimage,andasmallportionofresultingLPMspectrumisusedtocalculatethewatermarkpositions.Thistechniqueavoidscomputinginverselog-polarmapping(ILPM)topreservethequalityofthewatermarkedimage,andavoidsexhaustivesearchtosavecomputationtimeandreducefalsedetection.Experimentalresultsdemonstratethatthedigitalwatermarkingtechniqueisinvariantandrobusttorotation,scaling,andtranslationtransformation.

简介：Pointeranalysisisatechniquetoidentifyatcompile-timethepotentialvaluesofthepointerexpressionsinaprogram,whichpromisessignificantbenefitsforoptimizingandparallelizingcompilers.Inthispaper,anewapproachtopointeranalysisforassignmentsispresented.Inthisapproach,assignmentsareclassifiedintothreecategories:pointerassignments,structure(union)assignmentsandnormalassignmentswhichdon'taffectthepoint-toinformation.Pointeranalysesforthesethreekindsofassignmentsrespectivelymakeuptheintegratedalgorithm.Whenanalyzingapointerassignment,anewmethodcalledexpressionexpansionisusedtocalculateboththelefttargetsandtherighttargets.Theintegrationofrecursivedatastructureanalysisintopointeranalysisisasignificantoriginalityofthispaper,whichuniformsthepointeranalysisforheapvariablesandthepointeranalysisforstackvariables.ThisalgorithmisimplementedinAgassiz,ananalyzingtoolforCprogramsdevelopedbyInstituteofParallelProcessing,FudanUniversity.Itsaccuracyandeffectivenessareillustratedbyexperimentaldata.

简介：Astudyhasbeenconductedonthealgorithmofsolvinggeneralizedoptimalsetofdiscriminantvectorsinthispaper.Thispaperproposesananalyticalalgorithmofsolvinggeneralizedoptimalsetofdiscriminantvectorstheoreticallyforthefirsttime.Alotofcomputationtimecanbesavedbecauseallthegeneralizedoptimalsetsofdiscriminantvectorscanbeobtainedsimultaneouslywiththeproposedalgorithm,whileitneedsnoiterativeoperations.Theproposedalgorithmcanyieldamuchhigherrecognitionrate.Furthermore,theproposedalgorithmovercomestheshortcomingsofconventionalhumanfacerecognitionalgorithmswhichwereeffectiveforsmallsamplesizeproblemsonly.ThesestatementsaresupportedbythenumericalsimulationexperimentsonfacialdatabaseofORL.

简介：Inthispaper,twoapproachesareusedtosolvethePerspectiveThree-PointProblem(P3P):thesymboliccomputationapproachandthegeometricapproach.Inthesymboliccomputationapproach,weuseWu-Ritt'szerodecompositionalgorithmtogiveacompletetriangulardecompositionfortheP3Pequationsystem.ThisdecompositionprovidesthefirstcompleteanalyticalsolutiontotheP3Pproblem.Inthegeometricapproach,wegivesomepuregeometriccriteriaforthenumberofrealphysicalsolutions.Thecompletesolutionclassificationfortwospecialcaseswiththreeandfourparametersisalsogiven.

简介：Inthispaper,anewcoveringalgorithmcalledFCV1ispresented.FCV1comprisestwoalgorithms,oneofwhichisabletofastsearchforapartialruleandexcludethelargeportionofneggativeexamples,theotheralgorithmincorporatesthemoreoptimizedgreedyset-coveringalgorithm,andrunsonasmallportionoftrainingexamples.Hence,thetrainingprocessofFCV1ismuchfasterthanthatofAQ15.

简介：Thetheoryofparameterizedcomputationandcomplexityisarecentlydevelopedsubareaintheoreticalcomputerscience.Thetheoryisaimedatpracticallysolvingalargenumberofcomputationalproblemsthataretheoreticallyintractable.Thetheoryisbasedontheobservationthatmanyintractablecomputationalproblemsinpracticeareassociatedwithaparameterthatvarieswithinasmallormoderaterange:Therefore,bytakingtheadvantagesofthesmallparameters,manytheoreticallyintractableproblemscanbesolvedeffectivelyandpractically.Ontheotherhand,thetheoryofparameterizedcomputationandcomplexityhasalsoofferedpowerfultechniquesthatenableustoderivestrongcomputationallowerboundsformanycomputationalproblems,thusexplainingwhycertaintheoreticallytractableproblemscannotbesolvedeffectivelyandpractically.Thetheoryofparameterizedcomputationandcomplexityhasfoundwideapplicationsinareassuchasdatabasesystems,programminglanguages,networks,VLSIdesign,parallelanddistributedcomputing,computationalbiology,androbotics.Thissurveygivesanoverviewonthefundamentals,algorithms,techniques,andapplicationsdevelopedintheresearchofparameterizedcomputationandcomplexity.Wewillalsoreportthemostrecentadvancesandexcitements,anddiscussfurtherresearchdirectionsinthearea.

简介：ThispaperpresentsanewandsimpleschemetodescribetheconvexhullinR^d,whichonlyusesthreekindsofthefacesoftheconvexhull.i.e.,thed-1-faces,d-2-facesand0-faces.Thus,wedevelopandefficientnewalgorithmforconstructingtheconvexhullofafinitesetofpointsincrementally.Thisalgorithmemploysmuchlessstorageandtimethanthatofthepreviously-existingapproaches.Theanalysisoftherunniingtimeaswellasthestorageforthenewalgorithmisalsotheoreticallymade.Thealgorithmisoptimalintheworstcaseforevend.

简介：Inthispaperanewtext-independentspeakerverificationmethodGSMSVisproposedbasedonlikelihoodscorenormalization.Inthisnovelmethodaglobalspeakermodelisestablishedtorepresenttheuniversalfeaturesofspeechandnormalizethelikelihoodscore.Statisticalanalysisdemonstratesthatthisnormalizationmethodcanremovecommonfactorsofspeechandbringthedifferencesbetweenspeakersintoprominence.Asaresulttheequalerrorrateisdecreasedsignificantly,verificationprocedureisacceleratedandsystemadaptabilitytospeakingspeedisimproved.

简介：Thispaperproposesaheuristicstateminimizationalgorithm(HSM2)forfinitestatemachines(FSM).HSM2focusesonthegenerationandadjustmentoftheclosedcover.Firstaninitialclosedcoverisgeneratedbyheuristicallyselectingpropermaximalcompatiblestosatisfyallthecoveringandclosureconditions,andthenitisadjustedtobeaminimalornearminimalclosedcoverbyheuristicallyremovingrepeatedstates.Experimentalresultsshowthatthealgorithmisfasterandobtainsbetterorthesamesolutionscomparedwithconventionalmethods.

简介：人的姿势预言是为设计的一个关键因素；工作区的评估，在用虚拟人的虚拟环境。这个工作介绍一个新接口；为直接的人的虚拟环境优化了姿势预言(D-HOPP)在活动范围姿势必要的地方，来临到预言数字人的现实主义的活动范围姿势躯干的使用，手臂，；颈。在取决于什么类型任务正在被完成的地方，D-HOPP基于竞争，；人的姿势被不同人的性能测量管理。人的性能测量是一个基于物理的度量标准，例如精力或不快，；在优化明确的表达用作一个目标函数。这个问题作为单个目的的优化(SOO)被提出有单个性能测量的问题；作为multi-objective-optimization(哞)有多重联合性能测量的问题。我们使用联合排水量，在势能的变化，；肌与骨的不快是性能测量。D-HOPP与一个广泛的还直觉的用户接口被装备，；结果在交互虚拟环境被介绍。

简介：Inthispaper,anewmethod,namedasL-treematch,ispresentedforextractingdatafromcomplexdatasources.Firstly,basedondataextractionlogicpresentedinthiswork,anewdataextractionmodelisconstructedinwhichmodelcomponentsarestructurallycorrelatedviaageneralizedtemplate.Secondly,adatabase-populatingmechanismisbuilt,alongwithsomeobject-manipulatingoperationsneededforflexibledatabasedesign,tosupportdataextractionfromhugetextstream.Thirdly,top-downandbottom-upstrategiesarecombinedtodesignanewextractionalgorithmthatcanextractdatafromdatasourceswithoptional,unordered,nested,and/ornoisycomponents.Lastly,thismethodisappliedtoextractaccuratedatafrombiologicaldocumentsamountingto100GBforthefirstonlineintegratedbiologicaldatawarehouseofChina.

简介：