工程硕士学位论文写作范例StudyonthePavementMaterialofPerviousConcrete(申请清华大学工程硕士专业学位论文)培养单位：计算机科学与技术系工程领域：计算机技术申请人：李某指导教师：某某某教授联合指导教师：某某某高工二○○九年三月工程硕士学位论文写作说明李某关于学位论文使用授权的说明本人完全了解清华大学有关保留、使用学位论文的规定，即：清华大学拥有在著作权法规定范围内学位论文的使用权，其中包括：（1）已获学位的研究生必须按学校规定提交学位论文，学校可以采用影印、缩印或其他复制手段保存研究生上交的学位论文；（2）为教学和科研目的，学校可以将公开的学位论文作为资料在图书馆、资料室等场所供校内师生阅读，或在校园网上供校内师生浏览部分内容。本人保证遵守上述规定。作者签名：导师签名：日期：日期：摘要摘要情感是人类智能的重要方面。为建立和谐的人机交互环境，计算机自然需要具有理解情感和表达情感的能力。本文在声学层次上系统性地研究了情感的区分特征和感知特征，并提出了情感语音的叠加模型。论文的主要成果如下：1.分析了...，指出当前存在...问题。研究了...特点，提出了一种算法，并通过...实现了。2.提出了一种基于韵律强度的语音基频预测算法，....。3.设计了一个自学习的特征权值训练算法，提高了.....。4实现了....平台，验证了.....的有效性。关键词：关键词1关键词2关键词3关键词4关键词5IAbstractAbstractWiththerapiddevelopmentofinformationtechnology,computerbecomesanindispensabletoolinourdailylife.Tomakehuman-computerinteractionfriendlier,researchersofrelevantresearchfieldsapplythemselvesonthedevelopmentofnewhuman-computerinteractiontechnologies.Speech,asthemostnaturalwayinhumancommunication,isalsointhecenterofattention.AndtheHCI(human-computerinterface)technologies,whichbasedonspeechrecognition,speechsynthesis,andnaturallanguageunderstanding,havebeenrecognizedasthemostpromisingresearchdirection.Inrecentyears,asthedevelopmentofstatisticalmethodsforspeechsynthesis,largecorpusbasedText-to-Speech(TTS)systemhasbeenabletosynthesizehighqualityspeech.Butcomparedwithhumannaturalspeech,thesynthesizedspeechstillhassomeshortages,especiallyinprosodyexpression.Inthisthesis,speechprosodyinChinesePutonghuaisfirststudied,andaconclusionismadethatoneproblemwithcurrentprosodymodelingmethodsislackofaglobal-levelprosodyplanningprocess.Italsopointsoutthattheprosodicparametersforglobal-levelprosodyplanning,suchasprosodicstrength,isinneed.Inchapter2,anewprosodicstrengthestimationmethodbasedonParallelEncodingandTargetApproximation(PENTA)Modelisintroduced.Inthisestimationmethod,prosodicstrengthistakenasalatentvariableinphrase-levelprosodicplanningprocess,andprosodicstrengthfunction,whichisamappingfunctionbetweenacousticprosodicparametersandprosodicstrength,isrepresentedwithNeuralNetwork.Basedonthecorrelationbetweenprosodicstrengthandspeechunittargetcompletiondegree,prosodicstrengthfunctionsaretrainedautomaticallywithaspeechcorpus.Inchapter3,anF0generationmethodbasedonprosodicstrengthisproposed.InthisnewF0predictionmethod,theglobalprosodyplanningproblemistackledthroughaddingaglobalprosodicstrengthplanningprocessbeforepitchpredictionforspeechunits.Ithasbeenwidelyacceptedthat,inhumanspeechcommunicationthereisaprosodypre-planningprocessforeachprosodicphrasebeforearticulation,andIIIAbstractthensyllablesarearticulatedaccordingtotheirpre-plannedprosodicresults.Inthismethod,prosodicstrengthischosenasthelatentvariableforphraselevelprosodicplanning,andtheprosodyplanningprocessissimulatedwithprosodicstrengthmodeling.Soinprosodicprediction,aprosodicstrengthp...
