英語聲學(xué)和語音學(xué)英文版ppt課件市公開課一等獎百校聯(lián)賽優(yōu)質(zhì)課金獎名師賽課獲獎?wù)n件_第1頁
英語聲學(xué)和語音學(xué)英文版ppt課件市公開課一等獎百校聯(lián)賽優(yōu)質(zhì)課金獎名師賽課獲獎?wù)n件_第2頁
英語聲學(xué)和語音學(xué)英文版ppt課件市公開課一等獎百校聯(lián)賽優(yōu)質(zhì)課金獎名師賽課獲獎?wù)n件_第3頁
英語聲學(xué)和語音學(xué)英文版ppt課件市公開課一等獎百校聯(lián)賽優(yōu)質(zhì)課金獎名師賽課獲獎?wù)n件_第4頁
英語聲學(xué)和語音學(xué)英文版ppt課件市公開課一等獎百校聯(lián)賽優(yōu)質(zhì)課金獎名師賽課獲獎?wù)n件_第5頁
已閱讀5頁,還剩21頁未讀, 繼續(xù)免費閱讀

下載本文檔

版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請進行舉報或認(rèn)領(lǐng)

文檔簡介

SpeechacousticsandphoneticsLouisC.W.PolsInstituteofPhoneticSciences(IFA)AmsterdamCenterforLanguageandCommunication(ACLC)NATO-ASI“DynamicsofSpeechProductionandPerception”IlCiocco,Tuscany,Italy,July1,1/26OverviewDynamicsinspeechacousticsContourmodeling(mainlyformants)AspectsofspectralundershootModelingVandCreductionPhoneticknowledgefromspeechcorporaIFA,CGN,TIMIT,foundspeechConclusions2/26July1st,2Speechacousticsandphonetics,IlCiocco3/26July1st,3Speechacousticsandphonetics,IlCioccoDynamicsinspeechacousticsDynamicsisthenorm,notstationarityarticulatoryefficiencyDynamicsiseverywheregenerallynowordboundariesinspeechdeletionofwords,syllables,phonemes;insertionwithin/betweenwordcoarticulation/assimilationvowelandconsonantreductionAcousticmanifestationssegmentduration,F0,loudness,spectralquality4/26July1st,4Speechacousticsandphonetics,IlCioccoDynamicsisthenormThespeakerspeaksassloppilyasthelistenersallowhimtodoincommunicationcommunicativeefficiencyArticulatoryvs.perceptualefficiencydospectraltransitionsfacilitateorhamperperception?—>seeotherpresentationSpeakerflexibility;speakingstyle(clearvs.sloppy);speakingrate5/26July1st,5Speechacousticsandphonetics,IlCioccoDynamicsiseverywhereDeletion‘breadandbutter’/brEmbY3/‘Amsterdam’(Du)/Amst@rdAm/—>/Ams@dAm/‘koninklijke’(Du)/konI?kl@k@/—>/kol@k@/Insertionhomorganicglideinsertion:‘dieeen’(Du)/dij@n/Degemination‘iszichtbaar’(Du)/IszIxtbar/—>/IsIxbar/Reduction,coarticulation,assimilation6/26July1st,6Speechacousticsandphonetics,IlCioccoAcousticmanifestationspitch,loudness,formant,componentcontourscontourstylization(e.g.,pitchinpraat)contourmodelingn-thdegreecurvefitting (D.vanBergem)Legendrepolynomials ) (R.vanSon)16pointspersegment )(phoneme)segmentationbyhand(timeconsuming;non-consistent)automatically(viaforcedphonemerecognitionandapronunciationlexiconwithalternatives;systematicerrors)7/26July1st,7Speechacousticsandphonetics,IlCioccoContourmodelingallowsmodelingofspecificphenomenapitchaccentuation(vs.vowelonset)reduction,centralization,undershootallowsgenerationofstimuliforperc.expts.phonemeidentificationinextendingcontext2-alternativesforcedchoiceidentif.ofcontinuadiscrimination,RTallowsstatisticsonlargespeechcorporaTIMIT,CGN,IFA-corpus,Switchboard8/26July1st,8Speechacousticsandphonetics,IlCioccoStaticvs.dynamicVrecogn.seeWeenink()“VowelnormalizationswiththeTIMITacousticphoneticspeechcorpus”,IFAProc.24,117-123438males,bothtrain&testsent.ofTIMIT35,385vowelsegments,handsegmented13monophthongealvowelcategories1-Barkbandfilteranal.(18),intensity.normal.3framespersegment:centraland25msL/R9/26July1st,9Speechacousticsandphonetics,IlCioccoSomeresultsVowelclassif.(%)withdiscriminantfunctionsCondition#ItemsStatic1frameDynamic3framesOriginal35,385438x13x(1…25)59.366.9speakernormalized35,38562.269.2Vcentersperspeaker5,374438x1378.990.1speakernormalized5,37487.994.510/26July1st,10Speechacousticsandphonetics,IlCioccoFormanttracks/speakingratePh.D.thesisRobvanSon(1993)“Spectro-temporalfeaturesofvowelsegments”seealsoSpeechComm.13,135-148(Pols&vSon)850-wordstext,readatnormalandfastratehandsegmentationof7mostfreq.V+schwaformanttracksvia16pointspersegm.or5Legendrepolynomialsinfluenceofrate,V-dur.,context,sent.acc.evidenceforduration-controlledundershoot?11/26July1st,11Speechacousticsandphonetics,IlCioccoSomeresultsnodifferencesforF1/F2invowelcenterfornormal-orfast-ratespeech;onlysomeover-allriseinF1forfastrate(irrespectiveofV)sameformanttrackshape(normalizedto16points)fornormal-orfast-ratespeechsameresultswhenusingthemoreelaborateLegendrepolynomialsConcl.:changesinV-durationdonotchangetheamountofundershoot —>activecontrolofarticulationspeed12/26July1st,12Speechacousticsandphonetics,IlCioccoFormantrepresentationszerothorderLegendreLegendrepolynomialcoefficients(meanFiinvowelsegment)secondorderpolynomials(axesreversed)ee13/26July1st,13Speechacousticsandphonetics,IlCioccoModelingvowelreductionPh.D.thesisDickvanBergem(1995)“Acousticandlexicalvowelreduction”seealsoSpeechCommunication16,329-358lexicalVreductionFr/bet?/vs.Du/b@tOn/acousticVreduction/banan,bAnan,b@nan/f(sent.acc.,w.str.,w.class):can-candy-canteencoarticulatoryeffectsontheschwaC1@C2V-andVC1@C2-type

nonsensewordsperceptualeffects(fullVorschwa,f.i.‘a(chǎn)nanas’)14/26July1st,14Speechacousticsandphonetics,IlCioccoSomeresultsTheschwaisnotjustacentralizedvowelbutsomethingthatiscompletelyassimilatedwithitsphonemiccontextt-nw-l15/26July1st,15Speechacousticsandphonetics,IlCioccoModelingconsonantreductionSp.Comm.(1999)28,125-140(vSon&Pols)20min.speech,bothspontaneousandread2x791similarVCV;handsegmented5aspectsofVandCreductionrelatedtocoarticulation:F2slopedifferencesatCV-vs.VC-boundaries;F2locusequations(F2onsetvs.F2target)relatedtospeakingeffort:duration;spectralCOG(meanfreq.);V-Csoundenergydifferences16/26July1st,16Speechacousticsandphonetics,IlCioccoSomeresultsVmarkedlyreducedinspontaneousspeechlowerF2-slopediff.inspontaneousspeech —>decreaseinarticulationspeednosystematiceffectonF2locusequation;Vonsetsandtargetschangeinconcert—>anyVreductionmirroredbycomparablechangeinCspont.sp.:VandCshorter;lowerCOG—>decreaseinvocalandarticulatoryeffort17/26July1st,17Speechacousticsandphonetics,IlCioccoAccesstolargecorporamore,andmorerealistic,dataphoneticknowledgeviastatisticalanalysesf.i.highlyaccessibleIFA-corpus(free,SQL)see“StructureandaccessoftheopensourceIFA-corpus”,IFAProc.24,15-26(vSon&Pols)on-linehttp://www.fon.hum.uva.nl/IFAcorpus/4M/4Fspeakers,5.5hrsofspeechfrominformaltoread+sent.,words,syllables~50Kwordssegm.andlabeledatphonemelevel18/26July1st,18Speechacousticsandphonetics,IlCioccoSomeresultsspeech+annot.+metadata:relationalDBrealizationoffinaln,f.i.Du‘geven’/xev@(n)/Style#wrds/@n//@/All%/@n/Informal5,25013043050.3Retelling6,229132362495.2LFHFNarr.story14,453180372552334230Sentences14,97020334054337Pseudo-sent2,55462198177All43,4564591,2711,73036Read19/26July1st,19Speechacousticsandphonetics,IlCioccoSpokenDutchCorpus(CGN)10Mwords,1,000hrsofspeechvarietyofstyles,incl.telephonespeechadultDutchandFlemishspeakersforlinguisticandtechnologicalresearchseevariousLRECandICSLPpapers()seealsohttp://lands.let.kun.nl/cgn/home.htmfullytranscribed:orthogr.,POS,lemmaspartlytranscr.:phonemic,prosodic,syntactic20/26July1st,20Speechacousticsandphonetics,IlCioccoTIMITpopularDBinacousticphoneticsandASRalsotelephoneversion(NTIMIT)handsegmented&labeledatphonemelevel438males,192females(8dialectregions)10sent./sp.(2fixed,1phon.compact,7diverse)sa1:“Shehadherdarksuitingreasywashwaterallyear”includesseparatetestdata(112M,56F)e.g.Ph.DthesisX.Wang(1997)“IncorporatingknowledgeonsegmentaldurationinHMM-basedcontinuousspeechrecognition”21/26July1st,21Speechacousticsandphonetics,IlCioccoUsefulinfo:durationalvariabilityAdoptedfromWang(1998)normalrate=95

primarystress=104wordfinal=136utterancefinal=186overallaverage=95ms22/26July1st,22Speechacousticsandphonetics,IlCiocconormalizedphonedurationspeakingrateall3,696trainingsent.(sx+si)ofTIMITtrainingset023/26July1st,23Speechacousticsandphonetics,IlCiocco‘found’speechDARPA-LVSRcommunityratherambitiousBroadcastNews(BN),Sp.Comm.37()<’95WSJNABreadsp.1995Marketplace1996F0-F5,FXpartitioned19973hrstestunpartit.1998+nonEngl.speechalso<10xRTaudiotrainingdata100hrs10hrs55hrs+50hrs+100hrstext(forLM)430K122M540M>900Mbest%WERontestset27.0%27.1%1:46hrs16.2%3hrs13.5—>16.1%3hrs (10xRT)ForProc.DARPAWorkshops,see/speech/proc/darpa99/ind

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時也不承擔(dān)用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論