Improving short utterance based i-vector speaker recognition using source and utterance-duration normalization techniques

Biblos-e Archivo/Manakin Repository

Show simple item record

dc.contributor.author Kanagasundaram, Ahilan
dc.contributor.author Dean, David
dc.contributor.author González Domínguez, Javier
dc.contributor.author Sridharan, Sridha
dc.contributor.author Ramos, Daniel
dc.contributor.author González-Rodríguez, Joaquín
dc.contributor.other UAM. Departamento de Tecnología Electrónica y de las Comunicaciones es_ES
dc.date.accessioned 2015-01-12T15:36:59Z
dc.date.available 2015-01-12T15:36:59Z
dc.date.issued 2013-08
dc.identifier.citation 14th Annual Conference of the International Speech Communication Association. Ed. by F. Bimbot, C. Cerisara, C. Fougeron, G. Gravier, L. Lamel, F. Pellegrino, and P. Perrier. August 25-29, 2013 en_US
dc.identifier.issn 2308-457X
dc.identifier.uri http://hdl.handle.net/10486/663059
dc.description Proceedings of Interspeech 2013, Lyon (France) en_US
dc.description.abstract A significant amount of speech is typically required for speaker verification system development and evaluation, especially in the presence of large intersession variability. This paper introduces a source and utterance-duration normalized linear discriminant analysis (SUN-LDA) approaches to compensate session variability in short-utterance i-vector speaker verification systems. Two variations of SUN-LDA are proposed where normalization techniques are used to capture source variation from both short and full-length development i-vectors, one based upon pooling (SUN-LDA-pooled) and the other on concatenation (SUN-LDA-concat) across the duration and sourcedependent session variation. Both the SUN-LDA-pooled and SUN-LDA-concat techniques are shown to provide improvement over traditional LDA on NIST 08 truncated 10sec-10sec evaluation conditions, with the highest improvement obtained with the SUN-LDA-concat technique achieving a relative improvement of 8% in EER for mis-matched conditions and over 3% for matched conditions over traditional LDA approaches. en_US
dc.description.sponsorship This project was supported by the European Commission Marie Curie ITN ” Bayesian Biometrics for Forensics¨ (BBfor2) network and the Spanish Ministerio de Economia y Competitividad under the project TEC2012-37585-C02-01. en_US
dc.format.extent 5 pág. es_ES
dc.format.mimetype application/pdf en
dc.language.iso eng en
dc.publisher International Speech Communication Association en_US
dc.relation.ispartof Interspeech en_US
dc.rights © 2013 ISCA
dc.subject.other speaker verification en_US
dc.subject.other i-vector en_US
dc.subject.other total-variability en_US
dc.subject.other LDA en_US
dc.subject.other WCCN en_US
dc.title Improving short utterance based i-vector speaker recognition using source and utterance-duration normalization techniques en_US
dc.type conferenceObject en
dc.subject.eciencia Telecomunicaciones es_ES
dc.relation.publisherversion http://www.isca-speech.org/archive/interspeech_2013/i13_2465.html
dc.identifier.publicationfirstpage 2465
dc.identifier.publicationlastpage 2469
dc.relation.eventdate August 25-29, 2013 en_US
dc.relation.eventnumber 14
dc.relation.eventplace Lyon (France) en_US
dc.relation.eventtitle 14th Annual Conference of the International Speech Communication Association, INTERSPEECH 2013 en_US
dc.relation.projectID info:eu-repo/grantAgreement/EC/FP7/238803 en
dc.type.version info:eu-repo/semantics/publishedVersion en
dc.contributor.group Análisis y Tratamiento de Voz y Señales Biométricas (ING EPS-002) es_ES
dc.rights.accessRights openAccess en
dc.authorUAM González Domínguez, Javier (261826)


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record