基于小波變換的子帶自適應(yīng)濾波在信號(hào)提取中的應(yīng)用研究.pdf

1、東北大學(xué)博士學(xué)位論文基于小波變換的子帶自適應(yīng)濾波在信號(hào)提取中的應(yīng)用研究姓名：張威申請(qǐng)學(xué)位級(jí)別：博士專業(yè)：檢測(cè)技術(shù)與自動(dòng)化裝置指導(dǎo)教師：王旭20060601東北大學(xué)博士學(xué)位論文AbstmetStudyontheApplicationforSignalDetectionwithWaveletTransformBasedonSubbandAdaptiveFilterAbstractFilterbandshavemanyimportantapp

2、licationsinmanyfieldsOverthepastdecade，therehavebeenconsiderableadvancesinsubband(ormulti—rate)signalprocessingTheresearchontheoryanddesignoffilterbankshasb∞omepopularInthispaper，throughsystematicallysummarizingandanalyz

3、ingthedevelopmenthistoryandcurrentsituationofthesub—bandfilterbanksindomesticandabroad，integratingtherecentgreatachievementsinmodemtechnologyofsignalprocessing，weproposeanovelextractingmethodofsignalsubbandadap僦onfiltere

4、dalgorithmbasedonwavelettransformtWT)TheWaveletTransformation(WDhasexcellentcharacteristicsoftimefrequaneylocalizationInthispaperwewillcarefullyanalyzetheeffectofⅥ呵oilrestrainingmechanismofthesignalnoiseinterferenceTheMa

5、llatalgorithminⅥ叮willdividedi。gitalsignalsintoseveralorthonorrnalsub—frequenciesandlocalpositionsofsignalswillberepresentedindifferentwaveletsubspacesrespectivelyrepresentingastrapshapeinfrequencydomain，notoverlappingtog

6、etherTheanalyzingmethodofsubbandfrequencyisarepresentationoffrequencydomainlocalizationfunctioneffectivelysolvingthesubstrappartitionproblemin曲straptilterFrequencypartitionthroughWTistomakedigitalsignatmultiscalemathemat

7、icapproximationanditcanbeequaltotilterdigitalsignglusingmultiplesamplingratiotiltergroupItcomprisesofascalingfunctionlowpassfilterrelatedtoⅥ叮functionandwave&highpassfilterWhenwaveletdecomposingthesefiltershavenooverlappi

8、ng，butoverlaythewholefrequencyoftheinputsignalDuringthedecomposition，thetimeresolutionofsignalisreducedintohalfandahalfinformationislostaswell；forthefinitelengthserial，itslengthofsignalishalvedInthiswaysubstrapfilterhass

9、horterimpulseresponsethanfull—strapfilterDuringtheselfadaptationupdatesofsubband’sweights，differentstepscallbeapplied，whichwillresultinproducingaquickereonstringentsplecdthanthatoffullbandadaptationfilterArtificialneural

10、networks(ANNs)havealongdevelopmenthistoryandithasmadegreatprogressinregentyearsIt15quiteadeptatforecastingandprediction，patternrecognitionandcontrolapplicationMoreandmorestudiesonANNsbasedsignalprocessingmodelinghavebeen

11、reportedoverthelastdecadeLargenumbersofaplieationsshowthatANNshasSOmanyadvantagessuchasparallel—operation，distributedmemory，selfadaptability，errorpermissibilityandSOonSoitissuitableforsolvingthosesignalprocessingproblems