语音识别 外文翻译 外文文献 英文文献.docx
《语音识别 外文翻译 外文文献 英文文献.docx》由会员分享,可在线阅读,更多相关《语音识别 外文翻译 外文文献 英文文献.docx(14页珍藏版)》请在冰豆网上搜索。
SpeechRecognition
VictorZue,RonCole,&WayneWard
MITLaboratoryforComputerScience,Cambridge,Massachusetts,USAOregonGraduateInstituteofScience&Technology,Portland,Oregon,USACarnegieMellonUniversityPittsburgh,Pennsylvania,USA
1DefiningtheProblem
Speechrecognitionistheprocessofconvertinganacousticsignal,capturedbyamicrophoneoratelephone,toasetofwords.Therecognizedwordscanbethefinalresults,asforapplicationssuchascommands&control,dataentry,anddocumentpreparation.Theycanalsoserveastheinputtofurtherlinguisticprocessinginordertoachievespeechunderstanding,asubjectcoveredinsection.
Speechrecognitionsystemscanbecharacterizedbymanyparameters,someofthemoreimportantofwhichareshowninFigure.Anisolated-wordspeechrecognitionsystemrequiresthatthespeakerpausebrieflybetweenwords,whereasacontinuousspeechrecognitionsystemdoesnot.Spontaneous,orextemporaneouslygenerated,speechcontainsdisfluencies,andismuchmoredifficulttorecognizethanspeechreadfromscript.Somesystemsrequirespeakerenrollment—ausermustprovidesamplesofhisorherspeechbeforeusingthem,whereasothersystemsaresaidtobespeaker-independent,inthatnoenrollmentisnecessary.Someoftheotherparametersdependonthespecifictask.Recognitionisgenerallymoredifficultwhenvocabulariesarelargeorhavemanysimilar-soundingwords.Whenspeechisproducedinasequenceofwords,languagemodelsorartificialgrammarsareusedtorestrictthecombinationofwords.
Thesimplestlanguagemodelcanbespecifiedasafinite-statenetwork,wherethepermissiblewordsfollowingeachwordaregivenexplicitly.Moregenerallanguagemodelsapproximatingnaturallanguagearespecifiedintermsofacontext-sensitivegrammar.
Onepopularmeasureofthedifficultyofthetask,combiningthevocabularysizeandthe1languagemodel,isperplexity,looselydefinedasthegeometricmeanofthenumberofwordsthatcanfollowawordafterthelanguagemodelhasbeenapplied(seesectionforadiscussionoflanguagemodelingingeneralandperplexityinparticular).Finally,therearesomeexternalparametersthatcanaffectspeechrecognitionsystemperformance,includingthecharacteristicsoftheenvironmentalnoiseandthetypeandtheplacementofthemicrophone.
Speechrecognitionisadifficultproblem,largelybecauseofthemanysourcesofvariabilityassociatedwiththesignal.First,theacousticrealizationsofphonemes,thesmallestsoundunitsofwhichwordsarecomposed,arehighlydependentonthecontextinwhichtheyappear.Thesephoneticvariabilitiesareexemplifiedbytheacousticdifferencesofthephoneme,Atwordboundaries,contextualvariationscanbequitedramatic—makinggasshortagesoundlikegashshortageinAmericanEnglish,anddevoandaresoundlikedevandareinItalian.
Second,acousticvariabilitiescanresultfromchangesintheenvironmentaswellasinthepositionandcharacteristicsofthetransducer.Third,within-speakervariabilitiescanresultfromchangesinthespeaker'sphysicalandemotionalstate,speakingrate,orvoicequality.Finally,differencesinsociolinguisticbackground,dialect,andvocaltractsizeandshapecancontributetoacross-speakervariabilities.
Figureshowsthemajorcomponentsofatypicalspeechrecognitionsystem.Thedigitizedspeechsignalisfirsttransformedintoasetofusefulmeasurementsorfeaturesatafixedrate,2typicallyonceevery10—20msec(seesectionsand11.3forsignalrepresentationanddigitalsignalprocessing,respectively).Thesemeasurementsarethenusedtosearchforthemostlikelywordcandidate,makinguseofconstraintsimposedbytheacoustic,lexical,andlanguagemodels.Throughoutthisprocess,trainingdataareusedtodeterminethevaluesofthemodelparameters.
Speechrecognitionsystemsattempttomodelthesourcesofvariabilitydescribedaboveinseveralways.Atthelevelofsignalrepresentation,researchershavedevelopedrepresentationsthatemphasizeperceptuallyimportantspeaker-independentfeaturesofthesignal,andde-emphasizespeaker-dependentcharacteristics.Attheacousticphoneticlevel,speakervariabilityistypicallymodeledusingstatisticaltechniquesappliedtolargeamountsofdata.Speakeradaptationalgorithmshavealsobeendevelopedthatadaptspeaker-independentacousticmodelstothoseofthecurrentspeakerduringsystemuse,(seesection).Effectsoflinguisticcontextattheacousticphoneticlevelaretypicallyhandledbytrainingseparatemodelsforphonemesindifferentcontexts;thisiscalledcontextdependentacousticmodeling.
Wordlevelvariabilitycanbehandledbyallowingalternatepronunciationsofwordsinrepresentationsknownaspronunciationnetworks.Commonalternatepronunciationsofwords,aswellaseffectsofdialectandaccentarehan