盲源分离算法及应用研究

《盲源分离算法及应用研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、哈尔滨工程大学博士学位论文盲源分离算法及应用研究姓名：王卫华申请学位级别：博士专业：计算机应用技术指导教师：黄凤岗20090301哈尔滨T程大学博十学何论文提出了一种基于匹配追踪的稀疏源信号恢复算法。针对匹配追踪算法在盲源分离的稀疏源信号恢复问题中的具体应用，将匹配追踪算法加以改进，有效地提高了算法的性能。提出的算法在病态混合矩阵的情况下，可以减少匹配失败时所带来误差，具有良好的鲁棒性。关键词：盲源分离；排序不确定性；欠定盲分离；听觉场景分析；稀疏分量分析；匹配追踪肓源分离算法及应用研究AbstractB1indsourceseparation(BSS)isanewlypresente

2、dandlivelyareaofsignalprocessingdomaininrecentyears，whichestimatessourcesgivenmixedsignalswithoutpriorknowledgesuchassourcesnumber，locationandmixingprocess．BSSisapowerfultoolofdataanalysisandsignalprocessing，andithasbeenappliedinmanyareassuchasbiologicalandmedicinesignalprocessing，datamining，spe

3、echsignalprocessing，patternrecognition，wirelesscommunicationandSOon．Consideringitswideapplicationandabilityofdataprocessing，ResearchershavepaidagreatdealofattentiontoBSSandmadeagreatprogressinthepasttwentyyears．Thisthesisresearchestheseparationalgorithmsofspeechsignals．Firstly，abriefintroduction

4、ofthedevelopmenthistoryandcurrentresearchstatusaresummarized，andseveralapplicationsoftheBSSaregiven．Then，thisthesiSintroducesseveralkindsofimportantalgorithinsofBSS．ThisthesisdiscussesthetheoriesandalgorithmsofBSS，andespeciallyresearchestheunderdeterminedblindsourceseparationproblem．Theresearche

5、sconcentrateonthefollowingtopics：TheblindseparationproblemofconvolutivemixturesanddelayedmixturesiSdiscussedinthiSthesiS．Theseparationwhentherearemoresourcesthanmixturesisresearchedemphatically．ThemainworksofthiSthesiSareasfollows：Thepermutationindeterminationprobleminthefrequencydomainblindsour

6、ceseparationisdiscussed．Byutilizingthecharacteristicthatamplitudecorrelationbetweenneighborbinsofthesamesignalisbetterthanthatofdifferentsignals，animprovedmethodbasedontheamplitudecorrelationbetweenneighborbinstoeliminatethepermutationindeterminationiSproposed．Itisfeasibletoimplement哈尔滨T程大学博十学位论

7、文ICAalgorithmandpermutationmethodsimultaneously．Theunderdeterminedblindsourceseparationalgorithmsintime—frequencydomainareresearched．Anovelblindsourceseparationalgorithmbasedoncomputationalauditorysceneanalysis(CASA)ispropos