外文翻译---自适应维纳滤波方法的语音增强

《外文翻译---自适应维纳滤波方法的语音增强》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、附录ADAPTIVEWIENERFILTERINGAPPROACHFORSPEECHENHANCEMENTM.A.AbdEl-Fattah*,M.I.Dessouky,S.M.DiabandF.E.AbdEl-samie#DepartmentofElectronicsandElectricalcommunications,FacultyofElectronicEngineeringMenoufiaUniversity,Menouf,EgyptE-mails:*maro_zizo2010@yahoo.com,#fathi_sa

2、yed@yahoo.comABSTRACTThispaperproposestheapplicationoftheWienerfilterinanadaptivemannerinspeechenhancement.TheproposedadaptiveWienerfilterdependsontheadaptationofthefiltertransferfunctionfromsampletosamplebasedonthespeechsignalstatistics(meanandvariance).Theadaptiv

3、eWienerfilterisimplementedintimedomainratherthaninfrequencydomaintoaccommodateforthevaryingnatureofthespeechsignal.TheproposedmethodiscomparedtothetraditionalWienerfilterandspectralsubtractionmethodsandtheresultsrevealitssuperiority.Keywords:SpeechEnhancement,Spect

4、ralSubtraction,AdaptiveWienerFilter1INTRODUCTIONSpeechenhancementisoneofthemostimportanttopicsinspeechsignalprocessing.Severaltechniqueshavebeenproposedforthispurposelikethespectralsubtractionapproach,thesignalsubspaceapproach,adaptivenoisecancelingandtheiterativeW

5、ienerfilter[1-5].Theperformancesofthesetechniquesdependonqualityandintelligibilityoftheprocessedspeechsignal.Theimprovementofthespeechsignal-tonoiseratio(SNR)isthetargetofmosttechniques.Spectralsubtractionistheearliestmethodforenhancingspeechdegradedbyadditivenoise

6、[1].Thistechniqueestimatesthespectrumoftheclean(noise-free)signalbythesubtractionoftheestimatednoisemagnitudespectrumfromthenoisysignalmagnitudespectrumwhilekeepingthephasespectrumofthenoisysignal.Thedrawbackofthistechniqueistheresidualnoise.Anothertechniqueisasign

7、alsubspaceapproach[3].Itisusedforenhancingaspeechsignaldegradedbyuncorrelatedadditivenoiseorcolorednoise[6,7].Theideaofthisalgorithmisbasedonthefactthatthevectorspaceofthenoisysignalcanbedecomposedintoasignalplusnoisesubspaceandanorthogonalnoisesubspace.Processingi

8、sperformedonthevectorsinthesignalplusnoisesubspaceonly,whilethenoisesubspaceisremovedfirst.Decompositionofthevectorspaceofthenoisysignalisperform