到達方位角、時間延遲、載波頻率、起始方位角和多重弦波頻率等多維度參數估測，常被應用在信號源定位、多重輸入多重輸出及無線通訊頻道探測等信號處理領域。在此論文中，對多維度參數估測問題中的二維方位角及多維度諧波檢索為兩個特例，我們提出了一些高效能且適當複雜度基於最大似然法則的多維度參數估測演算法。這個新的演算法從粗略到精細反覆地估測信號藉由使用濾波處理對信號做適當的分群。為了便於實施最大似然法則估計，使用了Pincus的理論和一種叫做重要性取樣的蒙地卡羅方法來確定全局最佳最大似然解。因此，估測信號可以精確地被估測伴隨著適當的複雜度。此外，受估測的參數不需要額外的運算負擔就可自動達成配對。本論文應用大量電腦模擬與其他文獻所論述的方法做比較來進行演算法的效能驗證。

Jointestimationofmultidimensionalparameters,suchasdirection-of-arrival
(DOA),delay,carrierfrequency,direction-of-departure(DOD),andsinusoidalfre-
quency,arisesinvariousfacetsofsignalprocessingapplicationssuchassourcelocation,
multiple-input-multipleoutput(MIMO)systemsandwirelesschannelsounding.In
this thesis,weproposesomeefficacious,yetmoderatecomplexitymaximumlikelihood
(ML)-based algorithmsforthetwospecialcasesofmultidimensionalparameteresti-
mation problems,namelytwo-dimension(2-D)DOAandmultidimensionalharmonic
retrieval(MHR).
The newalgorithmsiterativelyestimatetheparametersinaroughtofinemanner,
intervenedwithfilteringprocesstoseparatethesignalsintoappropriategroups.To
facilitate implementationoftheMLestimation,thetheoremofPincusandaMonte
Carlo methodknownasimportancesampling(IS)areemployedtodeterminetheglobal
optimumMLsolution.Assuch,theparameterscanbepreciselyestimatedwithonly
moderatecomplexity.Moreover,theestimatedparametersareautomaticallypaired
together withoutextracomputationaloverhead.Conductedsimulationshowthatthe
new algorithmcanprovidesatisfactoryperformancecomparedwithpreviousworks.

[1] L.C.Godara,“Applicationofantennaarraystomobilecommunications,part
II:Beam-forminganddirection-of-arrivalconsiderations,” IEEE Proc., vol.85,
no. 8,pp.1195-1245,Aug.1997.
[2] D.NionandN.D.Sidiropoulos,“Tensoralgebraandmultidimensionalharmonic
retrievalinsignalprocessingforMIMOradar,” IEEE Trans.SignalProcess., vol.
58, no.11,pp.5693-5705,Nov.2010.
[3] J.LiuandX.Liu,“Aneigenvector-basedapproachformultidimensionalfre-
quency estimationwithimprovedidentifiability,” IEEE Trans.SignalProcess.,
vol.54,no.12,pp.4543-4556,Dec.2006.
[4] M.AkcakayaandA.Nehorai,“MIMOradarsensitivityanalysisfortargetdetec-
tion,” IEEE Trans.SignalProcess., vol.59,no.7,pp.3241-3250,July2011.
[5] N.Lehmann,E.Fishler,A.Haimovich,R.Blum,D.Chizhik,L.Cimini,andR.
Valenzuela,“EvaluationoftransmitdiversityinMIMO-radardirectionfinding,”
IEEE Trans.SignalProcess., vol.55,no.5,pp.2215-2225,May2007.
[6] A.L.Swindlehurst,“Timedelayandspatialsignatureestimationusingknown
asynchronoussignals,” IEEE Trans.SignalProcess., vol.46,pp.449-461,no.2,
Feb.1998.
[7] W.Zhi,C.C.Ko,andF.Chin,“Multi-hopMLbaseddelayandangleestimation
for multipathwidebandFHsignals,”in Proc.IEEEVehicularTechnologyConf.,
pp. 157-161,2004.
[8] M.C.Vanderveen,C.B.PapadiasandA.Paulraj,“Jointangleanddelayesti-
mation (JADE)formultipathsignalsarrivingatanantennaarray,” IEEE Com-
municationsLetters, vol.1,no.1,pp.12-14,Jan.1997.
[9] S.Wang,J.Caffery,andX.Zhou,“Analysisofajointspace-timeDOA/FOA
estimator usingMUSIC,”in Proc.IEEEInt'lSymposiumPersonal,Indoorand
Mobile RadioCommunications, pp. B138-B142,2001.
[10] A.J.vanderVeen,M.C.Vanderveen,andA.Paulraj,“Jointangleandde-
layestimation(JADE)usingshift-invarianceproperties,” IEEE Trans.Signal
Process., vol.46,no.2,pp.405-418,Feb.1998.
[11] M.C.Vanderveen,A.J.vanderVeen,andA.Paulraj,“Estimationofmultipath
parameters inwirelesscommunications,” IEEE Trans.SignalProcess., vol.46,
no. 3,pp.682-690,Mar.1998.
[12] M.Pesavento,A.B.Gershman,andM.Haardt,“Unitaryroot-MUSICwith
a real-valuedeigendecomposition:Atheoreticalandexperimentalperformance
study,” IEEE Trans.SignalProcess., vol.48,no.5,pp.1306-1314,May2000.
[13] A.J.vanderVeen,P.B.Ober,andE.F.Deprettere,“Azimuthandelevation
computation inhighresolutionDOAestimation,” IEEE Trans.SignalProcess.,
vol.40,no.7,pp.1828-1832,July1992.
[14] A.H.TewfikandW.Hong,“Ontheapplicationofuniformlineararraybearing
estimation techniquestouniformcirculararrays,” IEEE Trans.SignalProcess.,
vol.40,no.4,pp.1008-1011,Apr.1992.
[15] U.OktelandR.L.Moses,“High-resolution3-Ddirection-of-arrivaldetermi-
nation forurbanmobileradio,”in Proc.IEEEInt'lSymposiumAntennasand
Propagations, vol.45,pp.672-682,Apr.1997.
[16] M.D.ZoltowskiandC.P.Mathews,“Real-timefrequencyand2-Dangleestima-
tion withsub-nyquistspatio-temporalsampling,” IEEE Trans.SignalProcess.,
vol.42,no.10,pp.2781-2794,Oct.1994.
[17] M.D.Zoltowski,M.Haardt,andC.P.Mathews,“Closed-form2-Dangleestima-
tion withrectangulararraysinelementspaceorbeamspaceviaunitaryESPRIT,”
IEEE Trans.SignalProcess., vol.44,no.2,pp.316-328,Feb.1996.
[18] J.C.MosherandR.M.Leahy,“RecursiveMUSIC:aframeworkforEEGand
MEG sourcelocalization,” IEEE Trans.BiomedicalEngineering, vol.45,no.11,
pp. 1342-1354,Nov.1998.
[19] B.FriedlanderandA.J.Weiss,“Directionfindinginthepresenceofmutual
coupling,” in Proc.IEEEInt'lSymposiumAntennasandPropagations, vol.39,
no. 3,pp.273-284,Mar.1991.
[20] J.-D.Lin,W.-H.Fang,andJ.-T.Chen,“ConstrainedTSTMUSICforjoint
spatio-temporalchannelparameterestimationinDS/CDMA,” WirelessCom-
municationandMobileComputing, vol.5,pp.57-67,Jan.2005.
[21] A.L.SwindlehurstandP.Stoica,“Maximumlikelihoodmethodsinradararray
signal processing,” ProceedingsoftheIEEE, vol.86,no.2,pp.421-441,1998.
[22] B.Ottersten,M.Viberg,andT.Kailath,“AnalysisofsubspacefittingandML
techniquesforparameterestimationfromsensorarraydata,” IEEE Trans.Signal
Process., vol.40,no.3,pp.590-600,Mar.1992.
[23] J.-T.Chen,A.Paulraj,andU.Reddy,“Multichannelmaximum-likelihoodse-
quence estimation(MLSE)equalizerforGSMusingaparametricchannelmodel,”
IEEE Trans.Communications, vol.47,no.1,pp.53-63,Jan.1999.
[24] M.P.Clark,andL.L.Scharf,“Two-dimensionalmodalanalysisbasedonmax-
imumlikelihood,” IEEE Trans.SignalProcess., vol.42,no.6,pp.1443-52,Jun.
1994.
[25] S.M.Kay, FundamentalsofStatisticalSignalProcessing:EstimationTheory.
Prentice-Hall,1993.
[26] J.A.FesslerandA.O.Hero,“Space-alternatinggeneralizedexpectation-
maximization algorithm,” IEEE Trans.SignalProcess., vol.42,no.10,pp.2664-
2677, Oct.1994.
[27] I.ZiskindandM.Wax,“Maximumlikelihoodlocalizationofmultiplesources
byalternatingprojection,” IEEE Trans.Acoust,Speech,SignalProcess., vol.
ASSP-36, pp.1553-1560,Oct.1998.
[28] B.H.Fleury,M.Tschudin,R.Heddergott,D.Dahlhaus,andK.I.Pedersen,
“Channel parameterestimationinmobileradioenvironmentsusingtheSAGE
algorithm,” IEEE JournalonSelectedAreasinCommunications, vol.17,pp.
434-450, Mar.1999.
[29] H.Wang,S.Kay,andS.Saha“Animportancesamplingmaximumlikelihood
direction ofarrivalestimator,” IEEE Trans.SignalProcess., vol.56,no.10,pp.
5082-5092, Oct.2008.
[30] M.H.KalosandP.A.Whitlock, Monte CarloMethods. New York:JohnWiley
and Sons,1986.
[31] M.Pincus“Aclosedformsolutionforcertainprogrammingproblems,” Oper.
Res., vol.16,pp.690-694,May1968.
[32] R.O.Schmidt,“Multipleemitterlocationandsignalparameterestimation,” in
Proc.RADCSpectralEstimationWorkshop, Rome, pp.243-258,NY.1979.
[33] R.RoyandT.Kailath,“ESPRIT-Estimationofsignalparametersviarotational
invariancetechniques,” IEEE Trans.Acoustics,Speech,andSignalProcess., vol.
37, no.7,pp.984-995,July1989.
[34] L.Ge,T.Chen,andX.Huang,“Simultaneousfrequencyanddirectionestimation
from parallel-arraydata,” IEEE Proc.Radar,SonarNavig., vol.142,pp.6-10,
1995.
[35] C.-H.Lin,W.-H.FangK.-H.Wu,andJ.-D.Lin,“Afastalgorithmforjoint
two-dimensionaldirectionofarrivalandfrequencyestimationviahierarchical
space-time decomposition,” Signal Process., vol.90,no.1,pp.207-216,2010.
[36] C.-H.Lin,andW.-H.Fang,“Efficientmultidimensionalharmonicretrieval:A
hierarchicalsignalseparationframework,” IEEE SignalProcess.Letters, vol.20,
no. 5,pp.427-430,May2013.
[37] C.-H.Lin,andW.-H.Fang,“Efficientestimationofsignalparametersviaro-
tational invariancetechnique-basedalgorithmwithautomaticpairingfortwo-
dimensional angleandpolarisationestimationusingcrosseddipoles,” IET Signal
Process., vol.8,issue.4,pp.309-319,June2014.
[38] C.-H.Lin,“Efficientsubspace-basedalgorithmsformultidimensionalparame-
ter estimation:Ahierarchicalsignalseparationframework,”Ph.D.dissertation,
National TaiwanUniversityofScienceandTechnology,Taiwan,2013
[39] M.Haardt,C.Brunner,andJ.H.Nossek,“Efficienthigh-resolution3-Dchannel
sounding,” in Proc.IEEEInt'lConf.VehicularTechnology, pp. 164-168,May
1998.
[40] G.W.Lank,“Theoreticalaspectsofimportancesamplingappliedtofalse
alarms,” IEEE Trans.onInformationTheory, vol.29,pp.73-82,1983.
[41] G.F.OrsakandB.Aazhang,“Onthetheoryofimportancesamplingapplied
to analysisofdetectionsystems,” IEEE Trans.onCommunications, vol.37,pp.
332-339, 1989.
[42] S.Saha,‘Optimalestimationoflinearandnonlinearparametersingenerallinear
modelsbyimportancesampling,”Ph.D.dissertation,Univ.ofRhodeIsland,
Kingston, 2001.
[43] K.V.Mardia, Statistics ofDirectionalData. New York:Academic,1972,ch.2,3.
[44] B.C.Lovell,P.J.Kootsookos,andR.C.Williamson,“Thecircularnatureof
discrete-time frequencyestimates,”in Proc.IEEEInt'lConf.Acoustics,Speech,
and SignalProcess., pp. 3369-3372,1991.
[45] H.L.VanTrees, Optimum ArrayProcessing. Wiley-Interscience,2002.
[46] M.Wax,T.-J.Shan,andT.Kailath,“Spatio-temporalspectralanalysisbyeigen-
structure methods,” IEEE Trans.Acous.SpeechandSignalProcess., vol.32,no.
4, pp.817-827,1984.
[47] M.D.Zoltowski,M.Haardt,andC.P.Mathews,“Closed-form2-Dangleestima-
tion withrectangulararraysinelementspaceorbeamspaceviaunitaryESPRIT,”
IEEE Trans.SignalProcess., vol.44,no.2,pp.316-328,Feb.1996.
[48] W.Zhang,W.Liu,J.Wnag,andS.Wu,“Computationallyefficient2-DDOA
estimation foruniformrectangulararrays,” Multidim. Syst.SignalProcess., vol.
25, issue.4,pp.847-857,Oct.2014.
[49] J.E.FernandezdelRioandM.F.Catedra-Perez,“Thematrixpencilmethod
for two-dimensionaldirectionofarrivalestimationemployinganL-shapedarray,”
IEEE Trans.AntennasPropag., vol.45,no.11,pp.1693-1694,Nov.1997.
[50] N.Yilmazer,andT.K.Sarkar,“2-Dunitarymatrixpencilmethodforefficient
direction ofarrivalestimation,” Digital SignalProcess., vol.16,issue.6,pp.
767-781, Nov.2006.
[51] G.H.GolubandC.F.VanLoan, Matrix Computations. 3rd ed. JohnsHopkins
UniversityPress,1996
[52] C.-H.Lin,andW.-H.Fang,“Jointangleanddelayestimationinfrequencyhop-
ping systems,” IEEE Trans.AerospaceandElectronicSystems, vol.49,no.2,
pp. 1042-1056,Apr.2013.
[53] A.G.Jaffer,“Maximumlikelihooddirectionfindingforstochasticsources:Asep-
arable solution,”in Proc.IEEEInt'lConf.Acoustics,Speech,andSignalProcess.,
pp. 2893-2896,1998.
[54] M.WaxandT.Kailath,“Detectionofsignalsbyinformationtheoreticcriteria,”
IEEE Trans.AcousticsSpeech,SignalProcess., vol.33,no.2,pp.387-392,Apr.
1985.
[55] P.Stoica,“ImprovedsequentialMUSIC,” IEEE Trans.AerospaceandElectronic
Systems, pp. 1230-1239,Oct.1995.
[56] M.HaardtandJ.A.Nossek,“SimultaneousSchurdecompositionofseveralnon-
symmetric matricestoachieveautomaticpairinginmultidimensionalharmonic
retrievalproblems,” IEEE Trans.SignalProcess., vol.46,no.1,pp.161-169,
Jan. 1998.
[57] M.Haardt,F.Roemer,andG.DelGaldo,“Higher-orderSVD-basedsubspace
estimation toimprovetheparameterestimationaccuracyinmultidimensional
harmonic retrievalproblems,” IEEE Trans.SignalProcess., vol.56,no.7,pp.
3198-3213, July2008.
[58] X.LiuandN.D.Sidiropoulos,“Onconstantmodulusmultidimensionalharmonic
retrieval,”in Proc.IEEEInt'lConf.Acoustics,Speech,andSignalProcess., Orlando,
FL.pp.2977-2980,May2002.
[59] K.N.Mokios,N.D.Sidiropoulos,M.PesaventoandC.F.Mecklenbrぴauker,
“On 3-DHarmonicretrievalforwirelesscommunicaitonchannelsounding,”in
Proc.IEEEInt'lConf.Acoustics,Speech,andSignalProcess., Montreal,Quebec,
Canada, pp.89-92,May2004.
[60] M.Pesavento,C.F.Mecklenbrぴauker,andJ.F.Bぴohme,“Multidimensionalrank
reduction estimatorforparametricMIMOchannelmodels,” EURASIP J.Appl.
Signal Process., pp. 1354-1363,Sep.2004.
[61] J.L.Chen,H.GuandW.M.Su,“AnewmethodforjointDODandDOA
estimation inbistaticMIMOradar,” Signal Process., vol.91,issue.8,pp.1803-
1811, Aug.2011.
[62] B.Tang,J.Tang,Y.ZhangandZ.Zheng,“Maximumlikelihoodestimationof
DOD andDOAforbistaticMIMOradar,” Signal Process., vol.93,issue.5,pp.
1349-1357, May2013.
[63] M.Jin,G.LiaoandJ.Li,“JointDODandDOAestimationforbistaticMIMO
radar,” Signal Process., vol.89,issue.2,pp.244-251,Feb.2009.