语音识别-中英文献翻译.doc

语音识别-中英文献翻译.doc

ID:55307303

大小:154.50 KB

页数:14页

时间:2020-05-09

语音识别-中英文献翻译.doc_第1页
语音识别-中英文献翻译.doc_第2页
语音识别-中英文献翻译.doc_第3页
语音识别-中英文献翻译.doc_第4页
语音识别-中英文献翻译.doc_第5页
资源描述:

《语音识别-中英文献翻译.doc》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、SpeechRecognitionVictorZue,RonCole,&WayneWardMITLaboratoryforComputerScience,Cambridge,Massachusetts,USAOregonGraduateInstituteofScience&Technology,Portland,Oregon,USACarnegieMellonUniversity,Pittsburgh,Pennsylvania,USA1DefiningtheProblemSpeechrecogni

2、tionistheprocessofconvertinganacousticsignal,capturedbyamicrophoneoratelephone,toasetofwords.Therecognizedwordscanbethefinalresults,asforapplicationssuchascommands&control,dataentry,anddocumentpreparation.Theycanalsoserveastheinputtofurtherlinguisticp

3、rocessinginordertoachievespeechunderstanding,asubjectcoveredinsection.Speechrecognitionsystemscanbecharacterizedbymanyparameters,someofthemoreimportantofwhichareshowninFigure.Anisolated-wordspeechrecognitionsystemrequiresthatthespeakerpausebrieflybetw

4、eenwords,whereasacontinuousspeechrecognitionsystemdoesnot.Spontaneous,orextemporaneouslygenerated,speechcontainsdisfluencies,andismuchmoredifficulttorecognizethanspeechreadfromscript.Somesystemsrequirespeakerenrollment---ausermustprovidesamplesofhisor

5、herspeechbeforeusingthem,whereasothersystemsaresaidtobespeaker-independent,inthatnoenrollmentisnecessary.Someoftheotherparametersdependonthespecifictask.Recognitionisgenerallymoredifficultwhenvocabulariesarelargeorhavemanysimilar-soundingwords.Whenspe

6、echisproducedinasequenceofwords,languagemodelsorartificialgrammarsareusedtorestrictthecombinationofwords.Thesimplestlanguagemodelcanbespecifiedasafinite-statenetwork,wherethepermissiblewordsfollowingeachwordaregivenexplicitly.Moregenerallanguagemodels

7、approximatingnaturallanguagearespecifiedintermsofacontext-sensitivegrammar.Onepopularmeasureofthedifficultyofthetask,combiningthevocabularysizeandthelanguagemodel,isperplexity,looselydefinedasthegeometricmeanofthenumberofwordsthatcanfollowawordafterth

8、elanguagemodelhasbeenapplied(seesectionforadiscussionoflanguagemodelingingeneralandperplexityinparticular).Finally,therearesomeexternalparametersthatcanaffectspeechrecognitionsystemperformance,includingthecharacteristicsoftheenvironmentalnoise

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。