A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions

Probabilistic linear discriminant analysis (PLDA) is the leading method for computing scores in speaker recognition systems. The method models the vectors representing each audio sample as a sum of three terms: one that depends on the speaker identity, one that models the within-speaker variability,...

Descripción completa

Guardado en:
Detalles Bibliográficos
Publicado: 2018
Materias:
Probabilistic linear discriminant analysis
Speaker recognition
Cost functions
Discriminant analysis
Speech communication
Speech processing
Acoustic characteristic
Acoustic conditions
Joint modeling
Speaker recognition system
Speaker variability
Test condition
Speech recognition
Acceso en línea:https://bibliotecadigital.exactas.uba.ar/collection/paper/document/paper_2308457X_v2018-September_n_p82_Ferrer
http://hdl.handle.net/20.500.12110/paper_2308457X_v2018-September_n_p82_Ferrer
Aporte de:
Biblioteca Digital - Facultad de Ciencias Exactas y Naturales (UBA) de Universidad de Buenos Aires
id paper:paper_2308457X_v2018-September_n_p82_Ferrer
record_format dspace
spelling paper:paper_2308457X_v2018-September_n_p82_Ferrer2023-06-08T16:35:33Z A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions Probabilistic linear discriminant analysis Speaker recognition Cost functions Discriminant analysis Speech communication Speech processing Acoustic characteristic Acoustic conditions Joint modeling Probabilistic linear discriminant analysis Speaker recognition Speaker recognition system Speaker variability Test condition Speech recognition Probabilistic linear discriminant analysis (PLDA) is the leading method for computing scores in speaker recognition systems. The method models the vectors representing each audio sample as a sum of three terms: one that depends on the speaker identity, one that models the within-speaker variability, and one that models any remaining variability. The last two terms are assumed to be independent across samples. We recently proposed an extension of the PLDA method, which we termed Joint PLDA (JPLDA), where the second term is considered dependent on the type of nuisance condition present in the data (e.g., the language or channel). The proposed method led to significant gains for multilanguage speaker recognition when taking language as the nuisance condition. In this paper, we present a generalization of this approach that allows for multiple nuisance terms. We show results using language and several nuisance conditions describing the acoustic characteristics of the sample and demonstrate that jointly including all these factors in the model leads to better results than including only language or acoustic condition factors. Overall, we obtain relative improvements in detection cost function between 5% and 47% for various systems and test conditions with respect to standard PLDA approaches. © 2018 International Speech Communication Association. All rights reserved. 2018 https://bibliotecadigital.exactas.uba.ar/collection/paper/document/paper_2308457X_v2018-September_n_p82_Ferrer http://hdl.handle.net/20.500.12110/paper_2308457X_v2018-September_n_p82_Ferrer
institution Universidad de Buenos Aires
institution_str I-28
repository_str R-134
collection Biblioteca Digital - Facultad de Ciencias Exactas y Naturales (UBA)
topic Probabilistic linear discriminant analysis
Speaker recognition
Cost functions
Discriminant analysis
Speech communication
Speech processing
Acoustic characteristic
Acoustic conditions
Joint modeling
Probabilistic linear discriminant analysis
Speaker recognition
Speaker recognition system
Speaker variability
Test condition
Speech recognition
spellingShingle Probabilistic linear discriminant analysis
Speaker recognition
Cost functions
Discriminant analysis
Speech communication
Speech processing
Acoustic characteristic
Acoustic conditions
Joint modeling
Probabilistic linear discriminant analysis
Speaker recognition
Speaker recognition system
Speaker variability
Test condition
Speech recognition
A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions
topic_facet Probabilistic linear discriminant analysis
Speaker recognition
Cost functions
Discriminant analysis
Speech communication
Speech processing
Acoustic characteristic
Acoustic conditions
Joint modeling
Probabilistic linear discriminant analysis
Speaker recognition
Speaker recognition system
Speaker variability
Test condition
Speech recognition
description Probabilistic linear discriminant analysis (PLDA) is the leading method for computing scores in speaker recognition systems. The method models the vectors representing each audio sample as a sum of three terms: one that depends on the speaker identity, one that models the within-speaker variability, and one that models any remaining variability. The last two terms are assumed to be independent across samples. We recently proposed an extension of the PLDA method, which we termed Joint PLDA (JPLDA), where the second term is considered dependent on the type of nuisance condition present in the data (e.g., the language or channel). The proposed method led to significant gains for multilanguage speaker recognition when taking language as the nuisance condition. In this paper, we present a generalization of this approach that allows for multiple nuisance terms. We show results using language and several nuisance conditions describing the acoustic characteristics of the sample and demonstrate that jointly including all these factors in the model leads to better results than including only language or acoustic condition factors. Overall, we obtain relative improvements in detection cost function between 5% and 47% for various systems and test conditions with respect to standard PLDA approaches. © 2018 International Speech Communication Association. All rights reserved.
title A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions
title_short A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions
title_full A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions
title_fullStr A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions
title_full_unstemmed A generalization of PLDA for joint modeling of speaker identity and multiple nuisance conditions
title_sort generalization of plda for joint modeling of speaker identity and multiple nuisance conditions
publishDate 2018
url https://bibliotecadigital.exactas.uba.ar/collection/paper/document/paper_2308457X_v2018-September_n_p82_Ferrer
http://hdl.handle.net/20.500.12110/paper_2308457X_v2018-September_n_p82_Ferrer
_version_ 1768544340818264064

Ejemplares similares