[Zhang2011Final]

《[Zhang2011Final]》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、Thisarticlehasbeenacceptedforinclusioninafutureissueofthisjournal.Contentisfinalaspresented,withtheexceptionofpagination.IEEETRANSACTIONSONCOMMUNICATIONS,ACCEPTEDFORPUBLICATION1SpectrumSensingAlgorithmsviaFiniteRandomMatricesWenshengZhang,StudentMember,IEEE,GiuseppeAbreu,Se

2、niorMember,IEEE,MamikoInamori,Member,IEEE,andYukitoshiSanada,Member,IEEEAbstract—WeaddressthePrimaryUser(PU)detection(spec-fundamentaltoenableefficientautonomous(cognitive)utiliza-trumsensing)problem,relevanttocognitiveradio,fromafinitetionofspectrumbymultipleradios.Thisprobl

3、em,referredrandommatrixtheoretical(RMT)perspective.Specifically,wetoasspectrumsensingisthereforeofmajorimportanceinemployrecently-derivedclosed-formandexactexpressionsforthefieldofwirelesscommunicationsandatthecoreofthethedistributionofthestandardconditionnumber(SCN)ofuncorre

4、latedandsemi-correlatedrandomdualcentralWishartcognitiveradioparadigm[1]–[3].matricesoffinitesizesinthedesignHypothesis-Testingalgo-IthasbeenrecentlyshownthatspectrumsensingalgorithmsrithmstodetectthepresenceofPUsignals.Inparticular,twoofsuperiorperformanceandrobustnesscanbe

5、designedusingalgorithmsaredesigned,withbasisontheSCNdistributioninthetheeigenvaluesofWishartrandommatricesconstructedfromabsence(ℋ0)andinthepresence(ℋ1)ofPUsignals,respectively.DuetoaninherentpropertyoftheSCN’s,theℋ0testrequireschannelsamples[4]–[7].noestimationofSNRoranyot

6、herinformationonthePUIn[4],forinstance,spectrumsensingmethodswerepro-signal,whiletheℋ1testrequiresSNRonly.Furtherattractiveposedwhichrelyonthepropertythattheeingenvaluesofadvantagesofthenewtechniquesare:?)duetotheaccuracyof?×?Wishartrandommatrices,with(?,?)→∞andathefiniteSCN

7、distributions,superiorperformanceisachievedunderafinitenumberofsamples,comparedtoasymptoticconstantaspectratio?≜?/?,followtheMarchenko-PasturRMT-basedalternatives;?)sinceexpressionstomodeltheSCN(MP)law[8].TheMPlawestablishesthatthelargest(??)andstatisticsbothintheabsenceandp

8、resenceofPUsignalaresmallest(?1)eigenvaluesoflargematricesconverge–underused,thest