非高斯混響背景下的信號檢測.pdf

1、IIIIIllllIIIIIIII0Ilrlllilllrlllllllll、t1437605分類號UDC工學碩士學位論文密級編號非高斯混響背景下的信號檢測碩士研究生：劉竹青指導教師：田坦教授學位級別：工學碩士學科、專業(yè)：信號與信息處理所在單位：水聲工程學院論文提交日期：2008年2月論文答辯日期：2008年3月學位授予單位：哈爾濱工程大學ABSTRACTReverberationisoneofthebasicphysicalpheno

2、menainunderwateracousticIthasbeenalongtimesincebegintesearchitSufferingitsGaussianmechanismrestrict，conventionaltechniquesofunderwateracousticalsignalprocessingcannotproductidealperformanceforantireverberationprocessingT

3、hispaperbringsnon—GaussiansignalprocessingtechniquesintoantireverberationprocessingforactivesonarBasedonthis，theoptimaldetectionofdeterministicsignalagainstreverberationbackgroundisfoundsuccessfullyTheprobabilitydensityf

4、unction(PDF)ofreverberationtransientamplitudecanbefittedbvGaussianmixture(GM)modelwhereasthePDFofenvelopebyKdistributionmodelGMparameterscanbeestimatedbytheexpectationmaximizationiteration(EM)algorithmwhileK—distribution

5、bythemaximumlikelihoodalgorithmandthemomentsestimatealgorithmAtypicalgausslamzmgfllterisstudiedindetailsItiscalledUfilterwhichisbasedonthePDFanditsderivateAfterthegaussianizingfilterthenonGaussianityofreverberationwouldd

6、ecreasePSDofreverberationcanbefittedbyARmodel。ARparametersCanbeestimatedbvthemaximumlikelihoodestimation(MLE)andtheweightedleastsquareSestimation(WLSE)ThenwiththeARfilterbasedonARparameterestimation，reverberationcanbepre

7、whitenedwellUtilizingcouplingestimationforPSDandPDFparameters，combiningprewhiteningandgaussianizingmodulescorrelateprewhiteningwltngaussianizing(CPWG)filtersissetupTherelationshipbetweentheCPWGandtheRESTiSatSOstudiedKeyw