基于語音的多媒體場(chǎng)景標(biāo)記輔助系統(tǒng).pdf

1、上海大學(xué)碩士學(xué)位論文基于語音的多媒體場(chǎng)景標(biāo)記輔助系統(tǒng)姓名：楊慶濤申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：控制理論與控制工程指導(dǎo)教師：李昕20060101海人掌頓_：學(xué)位論文ABSTRACTThemainobjectiveoftheindexingprocessistoassignlabelstotheaudiovisualdatainordertodescribeitscontentAudioindexmgplaysakeyroleinthisproc

2、essInthispaper，aspeechbasedmanmachinelabelaidedsystemformediaassetmanagementispresentedThesystemrecogIlizesmephrasesspokenbytllehumanannotatorautomaticallyandassistshimtomarkupshotsofsubjectsinvideomedia，Thosephrasesares

3、egmented廳omshortsentencesandmodeledbytheellipticalbasisfunction(EBF)net、ⅣorksExperimemalresultsindicatematuleSpeechbasedlabelaidedsystemispracticalandlasgreatpmmiseformediaassetmaIlagementThispaperismainlyfocusonthesyste

4、mdeVelopmentinkeywordspotcingarldindexing1tproVidedanexcellentkey、VordspottingengineItdevelopedsomenewmethodsaIldalgori伽msinkc”Ⅳordsponinginthefleldofsportsvideostoac上lieVeahi曲detectionrateThemainworkisasfollows：1Thecons

5、tnJctionofrecognitionnetⅥ，orkforrecognitionenginebasedonANNmodelW毫tI豇neda11ANNmodelwilhsomekeywordsoften2Kevwords：usedinsportsVideoThekeywordrecognitionisbasedontherecognitionnetworkWbachievedgoodperfIormancewimthe印plica

6、tionofadaptiveweighteddata如sionalgori吐lrIlusedintheoutputdecisionThedeVelopmeInofspeechlabelsystemncanfindtheⅥ，AvEfileswhichcontainthepronunciationofspecifiedkeywordsItisaVocabularyChinesekeywordspottingaIldinde姬ngsystem