Repository logo
 
Publication

A hierarchical broad-class classification to enhance phoneme recognition

2009-08Conference paper Open access
Simple item page
datacite.subject.fosCiências Naturais::Ciências da Computação e da Informação
datacite.subject.fosEngenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática
datacite.subject.sdg03:Saúde de Qualidade
datacite.subject.sdg10:Reduzir as Desigualdades
datacite.subject.sdg11:Cidades e Comunidades Sustentáveis
dc.contributor.authorLopes, Carla, Alexandra Calado Lopes
dc.contributor.authorPerdigão, Fernando
dc.date.accessioned2025-05-30T11:01:53Z
dc.date.available2025-05-30T11:01:53Z
dc.date.issued2009-08
dc.description17th European Signal Processing Conference, EUSIPCO 2009, 24 August 2009 through 28 August 2009 - Code 91099
dc.description.abstractIn this paper a hierarchical classification of different levels of phonetic information is proposed in order to improve phone recognition. In this paradigm several intermediate classifiers give posterior probability predictions for broad phonetic classes, achieving phone detail in the last layer. Class membership probabilities are weighted and combined in order to get a more robust phoneme prediction. A method for finding the best set of weights is also proposed based on discriminative training in a hybrid MLP/HMM system. Experiments show that the use of broad-class information enhances phone recognition. Relative improvements of 8% in Correctness and 5% in Accuracy were achieved in phoneme recognition on the TIMIT database compared to a baseline system.eng
dc.description.sponsorshipCarla Lopes would like to thank the Portuguese foundation: Fundação para a Ciência e a Tecnologia for the PhD Grant (SFRH/BD/27966/2006).
dc.identifier.citationC. Lopes and F. Perdigão, "A hierarchical broad-class classification to enhance phoneme recognition," 2009 17th European Signal Processing Conference, Glasgow, UK, 2009, pp. 1760-1764.
dc.identifier.isbn978-161-7388-76-7
dc.identifier.issn2219-5491
dc.identifier.urihttp://hdl.handle.net/10400.8/13035
dc.language.isoeng
dc.peerreviewedyes
dc.publisherIEEE Canada
dc.relationDETECÇÃO DE EVENTOS ACÚSTICO-FONÉTICOS PARA RECONHECIMENTO AUTOMÁTICO DE FALA
dc.relation.hasversionhttps://ieeexplore.ieee.org/document/7077691
dc.rights.uriN/A
dc.subjectHidden Markov models
dc.subjectAbstracts
dc.subjectDatabases
dc.subjectEngines
dc.titleA hierarchical broad-class classification to enhance phoneme recognitioneng
dc.typeconference paper
dspace.entity.typePublication
oaire.awardTitleDETECÇÃO DE EVENTOS ACÚSTICO-FONÉTICOS PARA RECONHECIMENTO AUTOMÁTICO DE FALA
oaire.awardURIhttp://hdl.handle.net/10400.8/13033
oaire.citation.conferenceDate2009-08
oaire.citation.conferencePlaceGlasgow, Scotland
oaire.citation.endPage1764
oaire.citation.startPage1760
oaire.citation.titleEuropean Signal Processing Conference (EUSIPCO)
oaire.fundingStreamFARH
oaire.versionhttp://purl.org/coar/version/c_970fb48d4fbd8a85
person.familyNameLopes
person.givenNameCarla
person.identifier.ciencia-idAF14-3048-F510
person.identifier.orcid0000-0002-5366-0016
relation.isAuthorOfPublication4dfbaf0a-8c0b-4eaf-b1ad-0e590a5f3524
relation.isAuthorOfPublication.latestForDiscovery4dfbaf0a-8c0b-4eaf-b1ad-0e590a5f3524
relation.isProjectOfPublication4e8fdd81-def3-4535-bb9f-3b37e6bfc953
relation.isProjectOfPublication.latestForDiscovery4e8fdd81-def3-4535-bb9f-3b37e6bfc953

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
A hierarchical broad-class classification to enhance phoneme recognition.pdf
Size:
627.59 KB
Format:
Adobe Portable Document Format
Description:
In this paper a hierarchical classification of different levels of phonetic information is proposed in order to improve phone recognition. In this paradigm several intermediate classifiers give posterior probability predictions for broad phonetic classes, achieving phone detail in the last layer. Class membership probabilities are weighted and combined in order to get a more robust phoneme prediction. A method for finding the best set of weights is also proposed based on discriminative training in a hybrid MLP/HMM system. Experiments show that the use of broad-class information enhances phone recognition. Relative improvements of 8% in Correctness and 5% in Accuracy were achieved in phoneme recognition on the TIMIT database compared to a baseline system.
Download
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.32 KB
Format:
Item-specific license agreed upon to submission
Description:
Download

Collections

ESTG - Comunicações em conferências e congressos internacionais