188宝金博页面版

SPREADSPECTRUMINTERFERENCESUPPRESSIONUSINGADAPTIVETIME?FREQUENCYTILINGSBrianS?KrongoldMichaelL?KramerKannanRamchandranDouglasL?JonesCoordinatedScienceLaboratory?AdaptivUniversityofIllinoisdev????W?MainStreetonalUrbana?IL?????matcABSTRACTInterferencesuppressioninspreadspectrumcommunica?tionsystemsisoftenessentialforachievingmaximumsys?temperformance?Existinginterferencesuppressionmeth?odsdonotperformwellformosttypesofnonstation?aryinterference?We?rstconsiderinterferencesuppressionschemesbasedonadaptiveorthogonaltime?frequencyde?compositions?suchaswaveletpacketandarbitrarydyadictime?frequencytilings?Thesemethodsoftenreduceinter?ferencesubstantially?buttheirperformancecanvarydra?maticallywithminorchangesininterferencecharacteris?ticssuchasthecenterfrequency?Tocircumventthesedrawbacks?weproposeamultipleoverdeterminedtiling?MODT?withanaccompanyingblindinterferenceexci?sionschemewhichappearsverypromisingformitigatingtime?frequency?concentratedinterference?Simulationswithnarrowband?impulsive?andsimultaneousimpulsiveandnarrowbandinterferencecomparetheperformanceofthevariousmethodsandillustratethepromiseofapproachesbasedonmultipleoverdeterminedtilings???INTRODUCTIONDirectSequenceSpreadSpectrum?DS?SS?isaveryimpor?tantmethodofdatacommunication?Bysuperimposingapseudorandom?PN?sequenceoneachdatabit?thedataisspreadoveralargerbandwidthandislesssusceptibletointerfererswhilebeingmoresecure?Atthereceiver?thesig?nalissimplydespreadbacktoitsoriginalbandwidthandthedataisdemodulated?Ifaninterfererisremovedatthereceiverbeforedespread?ing?theperformanceoftheDS?SSsystemcanbegreatlyimproved?Anumberofdi?erentmethodsexistwhichworkwellforcertaintypesofinterferers?Theseincludetech?niquesbasedonadaptive?lteringandminimummean?square?error?MMSE?criteria????OverdeterminedFFT?basedmethodshavealsobeendevelopedfornarrowbandinterferencesuppression????TheFFT?basedandadaptiveMMSEtechniquesareverye?ectiveforstationary?narrowbandinterferersorwhentheinterfererisslowlyvarying?Iftheinterferersarehighlynonstationary?theseadaptivetechniquesdonotreactfastenoughtoremovethem?Techniquesbasedonlinearandbi?lineartime?frequencyrepresentations?TFRs?haverecentlybeendevelopedtoaddressthisproblem????Thebasicmoti?vationfortime?frequency?basedexcisionisthatmostinter?ferers?includingmostcommunicationssignals?otherman?madeinterfererssuchasradarorjammers?andevenimpul?sivenoisebursts?areinsomewayhighlyconcentratedintime?frequency?Atime?frequencydecompositionwillcon?centratesuchinterferersintoafewlargecoe?cients?anycoe?cientwhichexceedsathresholdisassumedtobedom?inatedbyinterferenceandissettozero?andaninterference?reducedsignalisthenreconstructedfromtheremainingco?e?cients?ThedesiredDS?SSsignalappearsnoise?likeandisspreadrelativelyevenlyoverallcoe?cientsandisthusmostlypreservedbytheinterferenceexcision?Existingtime?frequencymethodscanbee?ectiveforcer?taintypesofinterferers?buttheymaystrugglewithothertypesofnonstationaryinterferenceorwithmultipleinter?ferers?Waveletpackettransforms???canadaptanorthog?onalsubbanddecompositiontobettermatchsomeinterfer?ers?etime?frequencyexcision?hasrecentlybeenelopedin????E?ectively?thismethodadaptsanorthog?subbanddecompositionineithertimeorfrequencytohthesubbandstothefrequency?ortime?centeroftheinterference?Thisdecompositionissimilartowaveletpacketsexceptthatitallowsthree?bandaswellastwo?bandsplits?therebyoftenallowingbettermatchingtothecenterfrequencyoftheinterference?Moregeneralorthog?onaltime?frequencytilingshavebeendevelopedforbettermatchingabroaderclassofsignals????Unlikethewaveletpacketdecompositions?thegeneralmethodallowscthetilingpatterntochangeovertime?Inthispaper?weapplywaveletpacketandgeneralor?thogonaltime?frequencytilingstoDS?SSinterferenceexci?sion?We?ndthattheirperformanceisoftendisappointing?andarguethattheirweaknessstemsfromtheuseoforthog?onaltilings?Wepresentanewmethodforinterferenceexci?sionbasedonoverdeterminedtilings?Thesignalrepresen?tationisoversampledinbothtimeandfrequency?providingbetteralignmentwithinterfererswitharbitraryo?sets?andoverdeterminedintileshape?allowingreasonablematchingwithawidevarietyofinterferencecharacteristics???TIME?FREQUENCYTILINGSAone?dimensionalsignalcanberepresentedjointlyintermsofbothtimeandfrequencybyusingtime?frequencyorwaveletbases?Eachbasisfunctionofthisexpansionwillhavesomeregioninthetime?frequencyplanewheremostofitsenergyislocalized?Ifthisregionisde?nedtoberect?angularforeachbasisfunction?thenthesetofregionsisToappearinProc?ICASSP????Apr??????Munich?Germany??IEEE????