情感語(yǔ)音合成中的韻律研究.pdf

1、中國(guó)科學(xué)技術(shù)大學(xué)碩士學(xué)位論文情感語(yǔ)音合成中的韻律研究姓名：孫璐申請(qǐng)學(xué)位級(jí)別：碩士專(zhuān)業(yè)：信號(hào)與信息處理指導(dǎo)教師：王仁華20050501!里型蘭墊查叁蘭!堅(jiān)!：笙苧壘!!!!!璺AbstractWiththedevelopmentofcorpus—basedconcatinativespeechsynthesis，themachine—generatedspeechhasattainedhighfluencyandunderstandabi

2、lityandhasbeenusedinmanyareasofapplicationsHoweverasaresultofthelimitationsinspeechdatabaseconstructionandresearchesinsynthesizingmethods，currentlythesynthesizedspeechisgeneratedwithoutanyemotionorintonationvarietytheref

3、oreitsapplicationisseverelyrestricted，andgraduallybecamethebottleneckofthepopularizationofspeechsynthesis，ThereforespeechsynthesisbecomesthehotspotofspeechsynthesisresearchInthisthesis，thepresentedaretheacousticalcharact

4、eristicsofemotionalsynthesis，themethodsofprosodypredictionand0111“researchinemotionalspeechprosodypredictionGenerallNthisthesisiscomposedoffourparts：Thefirstchapterintroducestheconceptofemotionalspeech，discussestherecord

5、ingofemotionalspeechcorpusindetail，andlistsseveralemotionalspeechcorpusesasexamplesThen，thispartshowstheacousticalcharacteristicsofemotionalspeechwhicharesummarizedbyotherresearchers，andalsothepopularmethodsforemotionals

6、peechsynthesisInthesecondchaptegtheprosodypredictionmethodsarein廿oduced，includingrule—driyenmethod，data—drivenmethod，specificallythestatisticalmethod，theneuralnetworkmethod，data—miningmethod，andprosodicmodelingmethod，suc

7、hasTargetmodel，F(xiàn)ujisakimodel，SOPdurationmodelandSFCprosodymodelThethirdchapterpresentsemotionalspeechcorpusrecording，theprosodypredictionofemotionalspeech，andalsoaback—endsynthesizerTobemorespecific，wefirstlypresentandso

8、lvetheproblemofrecordingtextcollectionandcorpusspeechrecordingSecondlyarulebasedmethodisdesignedonthebasisofanemotionalspeechdatabasefromInstituteofLinguistics，ChinaAcademyofSocialSciencesThirdlythetheoriesandworkingproc

9、eduresofSFCprosodymodelisintroducedindetail，andaschemeofprosodymodelingforemotionalspeechisproposedforimprovementAlsoonthebasisofSFCmodelingofemotionalspeech，therelationshipofmodellayersisanalyzedandaneuralto—emotionaltr