基于遺傳算法的盲信號分離的研究與實現(xiàn).pdf

1、901116碩士研究寶掌姣論文新疆大學(xué)論文題目(中文)：基于遺傳算法的宮信號分離的研究與實現(xiàn)論文題毯(英文)；ResearchandRealizationOilGeneticAlgorithmsbasedBlindSignalSeparationmethods磷究生姓名：兜家茂學(xué)科、專業(yè)：通信與信息系統(tǒng)研究方向；網(wǎng)絡(luò)通信與信息系統(tǒng)導(dǎo)師姓名職稱：賴惠成副教授論文答辯日期：2006年6月日學(xué)位授予霹期：2006年月目AbstractAsac

2、lassofsignalprocessingtechnique，blindsignalseparation(BSS)isusedtorecovertheoriginalsignalsfrohatheobservedOneswithlittlepriorinformationaboutsourcesandmixturesystemSoitsresearchbecomesimportantboththeoreticallyandpracti

3、callyinthefieldsofbiomedical，speechrecognitioncommunication，etcBynowBSShasbeenatopicalsubjectinmodernsignalprocessingfields’AccordingtothedifierentmixturemodeIsofsources1inearBSScanbedividedintotwoclasses：instantaneousmi

4、xtureBSSandconvolutivemixtureBS甌毯目囊譬dissertation，someclassicalalgorithmsofrecentBSShavebeenstudiednemaincontributionsareasfollows：ReviewingtheresearchsituationofBSSoverrecentdecadessystematicallyandchoosingsomepopularBSS

5、algorithmstostudyandanalyzebysimulatedexperimentsTwocommondefectsarefoundedinBSSalgorithmsbasedonindependentcomponentanalysisfICA)：④Theselectionofnone—linearfunction(NLF)inICAdependsonthekurtosisoforiginalsignals，whichde

6、gradestheperformanceofseparationseriouslywhentheobservedsignalsarethemixtureofSuperGaussianandSubGaussiansignals(爹AstheoptimizationapproachinICAforsearchingseparationmatrix(orseparationmatrixes)，Gradienthasshortcomingsth

7、atconfigurationofinitialvalueandthelengthofpacewillmaketheseparationalgorithmsstrapintolocaloptimumvaluesInthisthesistwonewBSSalgorithmsarcproposedrespectivelybasedontwodifferentmixturemodels0neistheBSSbasedonkerneldensi

8、tyestimation(KD勘andgeneticalgorithm(GA)，theotheristheblinddeconvolutionbasedonhi曲ordercrosscumulantsandGAWithoutNLFtheperformanceofseparationinbothalgorithmsisindependentwiththekurtosisofthesourcesOntheotherhand，globalse

9、archingGAavoidsthedefectofgFadicntsothattherealoptimumvoluescanbefoundintheendThecorrectnessandvalidationofthetwoalgorithmsaretestifiedbytheoryanalysisandsimulatedexperimentsThesimulatedresultsalsoshowthattheproposedmeth