MAP Based Speaker Adaptation in Very Large Vocabulary Speech Recognition of Czech

Simple item page
dc.contributor.authorCerva, Petr
dc.contributor.authorNouza, Jan
dc.coverage.issue3cs
dc.coverage.volume13cs
dc.date.accessioned2016-04-26T11:04:13Z
dc.date.available2016-04-26T11:04:13Z
dc.date.issued2004-09cs
dc.description.abstractThe paper deals with the problem of efficient adaptation of speech recognition systems to individual users. The goal is to achieve better performance in specific applications where one known speaker is expected. In our approach we adopt the MAP (Maximum A Posteriori) method for this purpose. The MAP based formulae for the adaptation of the HMM (Hidden Markov Model) parameters are described. Several alternative versions of this method have been implemented and experimentally verified in two areas, first in the isolated-word recognition (IWR) task and later also in the large vocabulary continuous speech recognition (LVCSR) system, both developed for the Czech language. The results show that the word error rate (WER) can be reduced by more than 20% for a speaker who provides tens of words (in case of IWR) or tens of sentences (in case of LVCSR) for the adaptation. Recently, we have used the described methods in the design of two practical applications: voice dictation to a PC and automatic transcription of radio and TV news.en
dc.formattextcs
dc.format.extent42-46cs
dc.format.mimetypeapplication/pdfen
dc.identifier.citationRadioengineering. 2004, vol. 13, č. 3, s. 42-46. ISSN 1210-2512cs
dc.identifier.issn1210-2512
dc.identifier.urihttp://hdl.handle.net/11012/58059
dc.language.isoencs
dc.publisherSpolečnost pro radioelektronické inženýrstvícs
dc.relation.ispartofRadioengineeringcs
dc.relation.urihttp://www.radioeng.cz/fulltexts/2004/04_03_42_46.pdfcs
dc.rightsCreative Commons Attribution 3.0 Unported Licenseen
dc.rights.accessopenAccessen
dc.rights.urihttp://creativecommons.org/licenses/by/3.0/en
dc.subjectSpeech recognitionen
dc.subjectspeaker adaptationen
dc.subjectmaximum a posteriori methoden
dc.subjecthidden Markov modelsen
dc.titleMAP Based Speaker Adaptation in Very Large Vocabulary Speech Recognition of Czechen
dc.type.driverarticleen
dc.type.statusPeer-revieweden
dc.type.versionpublishedVersionen
eprints.affiliatedInstitution.facultyFakulta eletrotechniky a komunikačních technologiícs
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
04_03_42_46.pdf
Size:
418.47 KB
Format:
Adobe Portable Document Format
Description:
Download
Collections
2004/3