简介:VoIP(VoiceoverIP)isarapidlygrowingareawithgreatmarketpotential.Topromoteitforbothcommericalandresearchpurposes,aprototypeVoIPsystembasedonstate-of-the-artotorolacommunicationtechniqueshasbeendeveloped.ItisagatewaysystemintegratingaPBXandaVoIPmodule.AllcomponentsthatH.323definestosupprotVoIPareimplementedintheVoIPmodule,thoughinasimplifiedmanner,Asanembeddedsystem,thesystemfeaturesrealtimenessandtaskdistributiveness.Anumberofaddiitonaltechniquesareusedtoimprovetheperformance,includingnoisesuppression,zerocopy,andbufferstructureoptimization.Whenrefinedininteroperability,thesystemwillalsoreadilyserveasaproduct.
简介:Typesystemprovidesaprecisedescriptionofaprogramminglanguage.Thisisaprerequisitefortheimplementationanduseoflanguage.Italsoconductsmechanicalandtransparenttype-checkingonprogramstopreventtheoccurrenceofexecutionerrorduringtherunningofprograms.So,itcanbesaidthat,ontheonehand,typesystemworksasaformaltooltodomathematicalanalysisoflanguage;ontheotherhand,itisaformalmethodforrigorouslyandpreciselydesigningandimplementinglanguage.Inthispaper,somebasicconceptsoftypesystemarediscussedfirst.Andthen,theimplementationofagraph-rewriting-basedfunctionallanguage-SClean'stypesystemisgivenindetails.Itishopedthattheproposedmethodofusingandimplementingtypesystemisofpracticalusefulness.
简介:Thedistributedcomputersystemdescribedinthispaperisasetofcomputernodesinterconnectedinaninterconnectionnetworkviapacket-switchinginterfaces.Thenodescommunicatewitheachotherbymeansofmessage-passingprotocols.Thispaperpresentstheimplementationofrendezvousfacilitiesashigh-levelprimitivesprovidedbyaparallelprogramminglanguagetosupportinterprocesscommunicationandsynchronization.
简介:Inthispaper,anovelbiontetricidentificationsystemispresentedtoidentifyaperson'sidentitybyhis/herpalmprint.Incontrasttoexistingpalmprintsystemsforcriminalapplications,theproposedsystemtargetsatthecivilapplications,whichrequireidentifyingapersoninalargedatabasewithhighaccuracyinreal-time.Thesystemisconstitutedbyfourmajorcomponents:UserInterfaceModule,AcquisitionModule,RecognitionModuleandExternalModule.Morethan7,000palmprintimageshavebeencollectedtotesttheperformanceofthesystem.Thesystemcanidentify400palmswithalowfalseacceptancerate,0.02%,andahighgenuineacceptancerate,98.83%.Forverification,thesystemcanoperateatafalseacceptancerate,0.017%andafalserejectionrate,0.86%.Theexecutiontimeforthewholeprocessincludingimagecollection,preprocessing,featureextractionandmatchingislessthan1second.
简介:Wehavedevelopedafuzzyneuralexpertsystemthathastheprecisionandlearningabilityofaneuralnetwork.Knowledgeisacquiredfromdomainexpertsasfuzzyrulesandmembershipfunctions.Then,theyareconvertedintoaneuralnetworkwhichimplementsfuzzyinferencewithoutrulematching.Theneuralnetworkisappliedtoproblem-solvingandlearnsfromthedataobtainedduringoperationtoenhancetheaccuracy.Thelearningabilityoftheneuralnetworkmakesiteasytomodifythemembershipfunctionsdefinedbydomainexperts.Also,bymodifyingtheweightsofneuralnetworksadaptively,theproblemofbeliefpropagationinconventionalexpertsystemscanbesolvedeasily.Convertingtheneuralnetworkbackintofuzzyrulesandmembershipfunctionshelpsexplaintheinnerrepresentationandoperationoftheneuralnetwork.
简介:PROOSisadistributedoperatingsystemrunningonthecomputingnodesofmassivelyparallelprocessingcomputerDawning-1000.Itisanefficientandeasilyextendiblemicrokerneloperatingsystem.ItsupportstheIntelNXmessagepassinginterfaceforcommunication.
简介:ThispaperintroducesakindofsemanticcomputationandpresentshowtocombineitintoourChineseQuestion-Answering(QA)system.Basedontwokindsoflanguageresources,HownetandCilin,wepresentanapproachtocomputingthesimilarityandrelevancybetweenwords.Usingtheseresults,wecancalculatetherelevancybetweentwosentencesandthengettheoptimalanswerforthequeryinthesystem.ThecalculationadoptsquantitativemethodsandcanbeincorporatedintoQAsystemseasily,avoidingsomedifficultiesinconven-tionalNLP(NaturalLanguageProcessing)problems.Theexperimentsshowthattheresultsaresatisfactory.
简介:ThispaperintroducesaheterogeneousdistributeddatabasesystemcalledLSZsystem,whereLSZisanabbreviationofLiShizhen,anancientChinesemedicalscientist.LSZsystemadoptsclusterasdistributeddatabasenode(orsite).Eachclusterconsistsofoneofseveralmicrocomputersandoneserver.Tepaperdescribesitsbasicarchitectureandtheprototypeimplementation,whichincludesqueryprocessingandoptimization,transactionmanageranddatalanguagetranslation.ThesystemprovidesauniformretrieveandupdateuserinterfacethroughglobalrelationaldatalanguageGRDL.
简介:Thispaperexposestheconcurrentcharacteristicinthecommunicaionbetweenprocessorsinatransputerbasedsystem,andproposesanapproximateformulaforassessingthegranularityofdistributingtheparallelprocessesamongprocessors.ThereallymeasuredresultsforapipelinesortingprogramrunggingontheT-rackconfirmthesededucedresults.
简介:Tousereasoningknowledgeaccuratelyandefficiently,manyreasoningmethodshavebeenproposed.However,thedifferencesinformamongthemethodsmayobstructthesystematicalanalysisandharmoniousintegrationofthem.Inthispaper,acommonreasoningmodelJUM(JudgementModel)isintroduced.AccordingtoJUM,acommonknowledgerepresentationformisabstractedfromdifferentreasoningmethodsanditslimitationisreduced.WealsoproposeanalgorithmfortransformingonetypeofJUMsintoanother.Insomecases,thealgorithmcanbeusedtoresolvethekeyproblemofintegratingdifferenttypesofJUMinonesystem.Itispossiblethatanewarchitectureofknowledge-basedsystemcanberealizedunderJUM.
简介:RecentresearchonDistributedArtificialIntelligence(DAI)hasfocuseduponagents'interactioninMultiagentSystem.sThispaperpresentsatextunderstandingorientedmultiagentdynamicinteractiontestbed(TUMIT);thetheoreticframeworkbasedupongametheory,thefree-market-likesystemmarchitecture,andexperimentationonTUMIT.UnlikeotherDAItestbeds,TUMITviewsdifferenttextunderstanding(TU)methodsasdifferent“computationalresources”,andmakesagentschoosedifferentTUpathsandcomputationalresourcesaccordingtotheresouceinformationonthebulletinsintheirhostcomputer.Therefore,inTUMIT,taskallocationiswhollydistributed.ThismakesTUMITworklikea“freemarket”.Insuchasystem,agents'choicesandresourceloadmayoscillate.Itisshowntheoreticallyandexperimentallythatifagentsusemulti-levelof“historyinformation”,theirbehaviorwilltendtoconvergetoaNashequilibriumsituation;andthatifagentsuse“fecall-forget”strategyon“historyinformation”,theconvergencecanbeacceleratedandtheagentscanacclimatethemselvestochangedenvironment.ComparedwithotherDAItestbeds,TUMITismoredistributed,andtheagentsinTUMITaremoreadaptivetothedynamicenvironment.
简介:Inthispaper,ageometry-basedpointcloudreductionmethodisproposed,andareal-timemobileaugmentedrealitysystemisexploredforapplicationsinurbanenvironments.Weformulateanewobjectivefunctionwhichcombinesthepointreconstructionerrorsandconstraintsonspatialpointdistribution.Basedonthisformulation,amixedintegerprogrammingschemeisutilizedtosolvethepointsreductionproblem.Themobileaugmentedrealitysystemexploredinthispaperiscomposedoftheofttineandonlinestages.Attheofflinestage,webuildupthelocalizationdatabaseusingstructurefrommotionandcompressthepointcloudbytheproposedpointcloudreductionmethod.Whileattheonlinestage,wecomputethecameraposeinrealtimebycombininganimage-basedlocalizationalgorithmandacontinuousposetrackingalgorithm.Experimentalresultsonbenchmarkandrealdatashowthatcomparedwiththeexistingmethods,thisgeometry-basedpointcloudreductionmethodselectsapointcloudsubsetwhichhelpstheimage-basedlocalizationmethodtoachievehighersuccessrate.Also,theexperimentsconductedonamobileplatformshowthatthereducedpointcloudnotonlyreducesthetimeconsumingforinitializationandre-initialization,butalsomakesthememoryfootprintsmall,resultingascalableandreal-timemobileaugmentedrealitysystem.
简介:Systemadministratordealswithmanyproblems,ascomputingenvironmentbecomesincreasinglycomplex.Systemswithanabilitytorecognizesystemstatesandadapttoresolvetheseproblemsofferasolution.Muchexperienceandknowledgearerequiredtobuildaself-adaptivesystem.Self-adaptivesystemshaveinherentdifficulties.Thispaperproposesatechniquethatautomaticallygeneratesthecodefortheself-adaptivesystem.Thusthesystemiseasiertobuild.Self-adaptivesystemsofpreviousresearchrequiredhighsystemresourceusage.Incorrectoperationcouldbeinvokedbyexternalfactorssuchasviruses.Weproposeanimprovedself-adaptivesystemapproachandapplyittovideoconferencesystemandrobotsystem.Wecomparedthelinesofcode,thenumberofclassescreatedbythedevelopers.Wehaveconfirmedthisenhancedapproachtobeeffectiveinreducingthesedevelopmentmetrics.