Articulatory-feature based sequence kernel for high-level speaker verification

Shi Xiong Zhang; Man Wai Mak

doi:10.1109/ICMLC.2008.4620884

Articulatory-feature based sequence kernel for high-level speaker verification

Shi Xiong Zhang, Man Wai Mak

The Hong Kong Polytechnic University

Research output: Chapter in book / Conference proceeding › Conference article published in proceeding or book › Academic research › peer-review

Abstract

Research has shown that articulatory feature-based phoneticclass pronunciation models (AFCPMs) can capture the pronunciation characteristics of speakers. However, the scoring method used in AFCPMs does not explicitly use the discriminative information available in the training data. To harness this information, this paper proposes converting speaker models to supervectors by stacking the discrete densities in AFCPMs. An AF-kernel is constructed from the supervectors of target speakers, background speakers, and claimants. An AF-kernel based SVM is then trained to classify the supervectors. Results show that AF-kernel scoring is complementary to likelihood-ratio scoring, leading to better performance when the two scoring methods are combined.

Original language	English
Title of host publication	Proceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC
Pages	2799-2804
Number of pages	6
Volume	5
DOIs	https://doi.org/10.1109/ICMLC.2008.4620884
Publication status	Published - 25 Dec 2008
Event	7th International Conference on Machine Learning and Cybernetics, ICMLC - Kunming, China Duration: 12 Jul 2008 → 15 Jul 2008

Conference

Conference	7th International Conference on Machine Learning and Cybernetics, ICMLC
Country/Territory	China
City	Kunming
Period	12/07/08 → 15/07/08

Keywords

Articulatory features
Kernels
Pronunciation models
Speaker verification
SVM

ASJC Scopus subject areas

Artificial Intelligence
Human-Computer Interaction
Control and Systems Engineering

More information

10.1109/ICMLC.2008.4620884

Cite this

@inproceedings{acdfcc651a2345ebbe3f662a43464349,

title = "Articulatory-feature based sequence kernel for high-level speaker verification",

abstract = "Research has shown that articulatory feature-based phoneticclass pronunciation models (AFCPMs) can capture the pronunciation characteristics of speakers. However, the scoring method used in AFCPMs does not explicitly use the discriminative information available in the training data. To harness this information, this paper proposes converting speaker models to supervectors by stacking the discrete densities in AFCPMs. An AF-kernel is constructed from the supervectors of target speakers, background speakers, and claimants. An AF-kernel based SVM is then trained to classify the supervectors. Results show that AF-kernel scoring is complementary to likelihood-ratio scoring, leading to better performance when the two scoring methods are combined.",

keywords = "Articulatory features, Kernels, Pronunciation models, Speaker verification, SVM",

author = "Zhang, {Shi Xiong} and Mak, {Man Wai}",

year = "2008",

month = dec,

day = "25",

doi = "10.1109/ICMLC.2008.4620884",

language = "English",

isbn = "9781424420964",

volume = "5",

pages = "2799--2804",

booktitle = "Proceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC",

note = "7th International Conference on Machine Learning and Cybernetics, ICMLC ; Conference date: 12-07-2008 Through 15-07-2008",

}

Zhang, SX & Mak, MW 2008, Articulatory-feature based sequence kernel for high-level speaker verification. in Proceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC. vol. 5, 4620884, pp. 2799-2804, 7th International Conference on Machine Learning and Cybernetics, ICMLC, Kunming, China, 12/07/08. https://doi.org/10.1109/ICMLC.2008.4620884

Articulatory-feature based sequence kernel for high-level speaker verification. / Zhang, Shi Xiong; Mak, Man Wai.
Proceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC. Vol. 5 2008. p. 2799-2804 4620884.

Research output: Chapter in book / Conference proceeding › Conference article published in proceeding or book › Academic research › peer-review

TY - GEN

T1 - Articulatory-feature based sequence kernel for high-level speaker verification

AU - Zhang, Shi Xiong

AU - Mak, Man Wai

PY - 2008/12/25

Y1 - 2008/12/25

N2 - Research has shown that articulatory feature-based phoneticclass pronunciation models (AFCPMs) can capture the pronunciation characteristics of speakers. However, the scoring method used in AFCPMs does not explicitly use the discriminative information available in the training data. To harness this information, this paper proposes converting speaker models to supervectors by stacking the discrete densities in AFCPMs. An AF-kernel is constructed from the supervectors of target speakers, background speakers, and claimants. An AF-kernel based SVM is then trained to classify the supervectors. Results show that AF-kernel scoring is complementary to likelihood-ratio scoring, leading to better performance when the two scoring methods are combined.

AB - Research has shown that articulatory feature-based phoneticclass pronunciation models (AFCPMs) can capture the pronunciation characteristics of speakers. However, the scoring method used in AFCPMs does not explicitly use the discriminative information available in the training data. To harness this information, this paper proposes converting speaker models to supervectors by stacking the discrete densities in AFCPMs. An AF-kernel is constructed from the supervectors of target speakers, background speakers, and claimants. An AF-kernel based SVM is then trained to classify the supervectors. Results show that AF-kernel scoring is complementary to likelihood-ratio scoring, leading to better performance when the two scoring methods are combined.

KW - Articulatory features

KW - Kernels

KW - Pronunciation models

KW - Speaker verification

KW - SVM

UR - http://www.scopus.com/inward/record.url?scp=57849094579&partnerID=8YFLogxK

U2 - 10.1109/ICMLC.2008.4620884

DO - 10.1109/ICMLC.2008.4620884

M3 - Conference article published in proceeding or book

SN - 9781424420964

VL - 5

SP - 2799

EP - 2804

BT - Proceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC

T2 - 7th International Conference on Machine Learning and Cybernetics, ICMLC

Y2 - 12 July 2008 through 15 July 2008

ER -

Articulatory-feature based sequence kernel for high-level speaker verification

Abstract

Conference

Keywords

ASJC Scopus subject areas

More information

Other files and links

Fingerprint

Cite this