增强的基于vq算法的说话人语音识别外文文献翻译

《增强的基于vq算法的说话人语音识别外文文献翻译》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、毕业设计（论文）外文翻译外文题目：EnhancedVQ-basedAlgorithmsforSpeechIndependentSpeakerIdentification译文题目：增强的基于VQ算法的说话人语音识别文献出处「《Audio-andVideo-basedBiometriePersonAuthentication,IntcrnationalConference,Avbpa,Gu订dford,Uk,Junc》，2003,2688:470-477外文作者：NingpingFan,JustinianRose且字数统计：英文1869单词，9

2、708字符；中文3008汉字外文文献:EnhancedVQ-basedAlgorithmsforSpeechIndependentSpeakerIdentificationAbstractWeighteddistancemeasureanddiscriminativetrainingaretwodifferentapproachestoenhanceVQ-basedsolutionsforspeakeridentification.ToaccountforvaryingimportanceoftheLPCcoefficientsinSV,t

3、heso-calledpartitionnormalizeddistancemeasuresuccessfullyusednormalizedfeaturecomponents.Thispaperintroducesanalternative,calledheuristicweighteddistance,toliftuphigherorderMFCCfeaturevectorcomponentsusingalinearformula.Thenitproposestwonewalgorithmscombiningtheheuristicwe

4、ightingandthepartitionnormalizeddistancemeasurewithgroupvectorquantizationdiscriminativetrainingtotakeadvantageofbothapproaches.ExperimentsusingtheTIMITcorpussuggestthatthenewcombinedapproachissuperiortocurrentVQ-basedsolutions(50%errorreduction).ItalsooutperformstheGaussi

5、anMixtureModelusingtheWaveletfeaturestestedinasimilarsetting・LIntroductionVectorquantization(VQ)basedclassificationalgorithmsplayanimportantroleinspeechindependentspeakeridentification(SI)systems.Althoughinbaselineform,theVQ-basedsolutionislessaccuratethantheGaussianMixtur

6、eModel(GMM),itofferssimplicityincomputation.Foralargedatabaseofoverhundredsorthousandsofspeakers,bothaccuracyandspeedareimportantissues.HerewediscussVQenhancementsaimedataccuracyandfastcomputation.1>1VQBasedSpeakerIdentificationSystemFig.1showstheVQbasedspeakeridentificati

7、onsystem.Itcontainsanofflinetrainingsub-systemtoproduceVQcodebooksandanonlinetestingsub-systemtogenerateidentificationdecision.Bothsub-systemscontainapreprocessingorfeatureextractionmoduletoconvertanaudioutteranceintoasetoffeaturevectors.Featuresofinterestintherecentlitera

8、turesincludetheMel-frequencycepstralcoefficients(MFCC),theLinespectrapairs(LSP),theWavele