简介:Auditorymodelhasbeenappliedtoseveralaspectsofspeechsignalprocessingfield,andappearstobeeffectiveinperformance.Thispaperpresentstheinversetransformofeachstageofonewidelyusedauditorymodel.Firstofallitisnecessarytoinvertcorrelogramandreconstructphaseinformationbyrepetitiousiterationsinordertogetauditory-nervefiringrate.ThenextstepistoobtainthenegativepartsofthesignalviathereverseprocessoftheHWR(HalfWaveRectification).Finallythefunctionsofinnerhaircell/synapsemodelandGammatonefiltershavetobeinverted.Thusthewholeauditorymodelinversionhasbeenachieved.Anapplicationofnoisyspeechenhancementbasedonauditorymodelinversionalgorithmisproposed.Manyexperimentsshowthatthismethodiseffectiveinreducingnoise.EspeciallywhenSNRofnoisyspeechislowitismoreeffectivethanothermethods.Thusthisauditorymodelinversionmethodgiveninthispaperisapplicabletospeechenhancementfield.
简介:Inordertoprovidethebasisforparameterselectionofvocaldiseasesclassification,anonlineardynamicmodelingmethodisproposed.Abiomechanicalmodelofvocalcordswithpolyporparalysis,whichcouplestoglottalairflowtoproducelaryngealsoundsource,isintroduced.Andthenthefundamentalfrequencyanditsperturbationparametersaresolved.Poincaresectionandbifurcationdiagramareappliedtononlinearanalysisofmodelvibration.Bychangingthepathologicalparametersorsubglottalpressure,thechangesoffundamentalfrequencyandLyapunovexponentsareanalyzed.Thesimulationresultsshowthat,vocalcordparalysisreducesthefundamentalfrequency,andthechaosoccursonlywithinacertainpressurerange;whilevocalcordwithapolypdon’treducethefundamentalfrequency,chaosdistributesthroughouttheentirerangeofpressure.Thereforethisstudyishelpfulforclassificationofpolypandparalysisbytheacousticdiagnoses.
简介:Arrayprocessingistoprocessthesignalscarriedbythepropagatingwavesreceivedatanarrayofsensors.Whenthesignalspropagatethroughthepracticalrandomtime-variantmedium,theirwavefrontscanshowtheprogres-sivelossesofcoherencewithincreasingspatialseparation.Thesedecorrelationsofwavefrontsresultinanangularspreadinthewavenumberspectrumcenteredaboutthetruesignaldircction-of-arrival.Thispaperputstheemphasisuponthearrayprocessingoftheangular-spreadsignalwhichiscalledtheGeneral-izedDirectional(GD)signalandaimstomatcharrayprocessingtothissignalmodelintheenergysense.Inthispaper,wealsopresentamethodofthecom-putersimulationorthegeneralizeddirectionalsignalmodel.Someresultsofcomputersimulationexperimentsandlake-tcstsinXinanjiangRiveraregiven.
简介:一个修改Parzen窗户方法,在低频率使分辨率高并且把光滑放在高频率,被建议获得统计模型。然后,当长句子被处理时,利用统计模型的一个性分类方法被建议,它有性分类的98%精确性。由男声音和女性表示的分离,与不同情感训练样品的讲话的平均数和标准差被用来创造相应情感模型。然后在测试样品和沥青的统计模型之间的Bhattacharyya距离,在沥青的speech.The正规化为情感识别被利用因为男声音和女声音也被考虑,以便说明他们直到一个一致空格。最后,讲话情感识别实验基于K最近的邻居显示出那,81%的正确的率被完成,在它仅仅是73.85%if的地方,传统的参数被利用。
简介:Aspeechsignalprocessingandfeaturesextractingmethodbasedoncomputationalauditorymodelisproposed.Thecomputationalmodelisbasedonpsychological,physiologicalknowledgeanddigitalsignalprocessingmethods.Ineachstageofahearingperceptionsystem,thereisacorrespondingcomputationalmodeltosimulateitsfunction.Basedonthismodel,speechfeaturesareextracted.Ineachstage,thefeaturesindifferentkindsoflevelareextracted.Afurtherprocessingforprimaryauditoryspectrumbasedonlateralinhibitionisproposedtoextractmuchmorerobustspeechfeatures.Allthesefeaturescanberegardedastheinternalrepresentationsofspeechstimulationinhearingsystem.Therobustspeechrecognitionexperimentsareconductedtotesttherobustnessofthefeatures.Resultsshowthattherepresentationsbasedontheproposedcomputationalauditorymodelarerobustrepresentationsforspeechsignals.
简介:AnewmethodologyofvoiceconversionincepstrumeigenspacebasedonstructuredGaussianmixturemodelisproposedfornon-parallelcorporawithoutjointtraining.Foreachspeaker,thecepstrumfeaturesofspeechareextracted,andmappedtotheeigenspacewhichisformedbyeigenvectorsofitsscattermatrix,therebytheStructuredGaussianMixtureModelintheEigenSpace(SGMM-ES)istrained.Thesourceandtargetspeaker’sSGMM-ESarematchedbasedonAcousticUniversalStructure(AUS)principletoachievespectrumtransformfunction.Experimentalresultsshowthespeakeridentificationrateofconversionspeechachieves95.25%,andthevalueofaveragecepstrumdistortionis1.25whichis0.8%and7.3%higherthantheperformanceofSGMMmethodrespectively.ABXandMOSevaluationsindicatetheconversionperformanceisquiteclosetothetraditionalmethodundertheparallelcorporacondition.TheresultsshowtheeigenspacebasedstructuredGaussianmixturemodelforvoiceconversionunderthenon-parallelcorporaiseffective.
简介:Acoupled-modesoundpropagationmodelwithcomplexeffectivedepthispresented,inordertoinvolvetheeffectofbranchlineintegralforacousticfieldinarange-dependentwaveguide.Theequationsofmotionandcontinuityareusedtoobtainthecoupledequations,whichsatisfyboundaryconditionsinthewaveguidewithvaryingtopographyandcontainonecouplingmatrix.Meanwhile,thecouplingsbetweendiscreteandcontinuousspectrumaredealtwithbasedoncomplexeffectivedepththeory.Numericalsimulationsshowthattheaccuracyoftransmissionlossisimprovedbythecoupledmodemodelwheneigenvaluesoftrappedmodesarelocatednearthebranchpoint.Theacousticfieldinanon-horizontallystratifiedwaveguidecanbecalculatedefficientlyandaccuratelybythismodel,andtheenergycorrespondingtotrappedmodes,leakymodesandbranchlineintegralcanbeconsideredadequately.
简介:在连续语音识别的音调模型集成的一个歧视的框架被建议。放大隐藏的Markov模型的可能性基于的方法使用模型依赖者重量光谱特征和音调模型基于音调的features.The重量是有区别地由最小的电话错误标准的trahined。模型重量的更改方程基于扩大Baum-Welch算法被导出。模型重量联合的各种各样的计划被评估,一种变光滑的技术被介绍成为训练对柔韧在适合上。建议方法是音调的音节输出和字符输出语音识别任务上的ewluated。试验性的结果证明建议方法由于给定的模型的更好的插值在二项任务比全球重量获得了9.5%和4.7%相对错误减小。这为音调模型集成证明歧视的训练模型重量的有效性。
简介:Time-frequencyanalysisiscombinedwitharrayprocessingtodevelopadirectionofarrival(DOA)estimationmethod.Thearraydatamodelisconstructedintime-frequencydomainbycrosstime-frequencydistributionbetweentheoutputofareferencesensorandthoseoftwosymmetricsub-arrays.Accordinglyasubspacemethodispresentedbasedontheaverageoftwosub-arrays'time-frequencydatavectormodelinsteadoftheconventionalarraymodel,toestimateDOAsofmultiplesignals.Becausethearraydataisprocessedbothinspatialdomainand2-Dtime-frequencydomain,theproposedmethodhasanabilitytoselectthesignalofinteresting,andissuitablefornon-stationarysignal.Additionally,themethodisrobusttonoiseandholdsanadvantageoflowcomputationalload.Simulationsareconductedtoverifytheefficiencyofthemethodandcomparisionismadewithothermethods.
简介: