Parameterization of prosodic feature distributions for SVM modeling in speaker recognition

Luciana Ferrer, Elizabeth Shriberg, Sachin Kajarekar, Mustafa (Kemal) Sonmez

Research output: Chapter in Book/Report/Conference proceedingConference contribution

24 Citations (Scopus)

Abstract

Multiple recent studies have shown that speaker recognition performance using frame-based cepstral features is improved by adding higher-level information, including prosodie and lexical features. This paper explores the important question of finding a good kernel for a system that models syllable-based prosodie features using support vector machines (SVMs). The system has been the best performing of our high-level systems in the last two NIST evaluations, and gives significant improvements when combined with cepstral-based systems. We introduce two new methods for transforming the syllable-level features into a single high-dimensional vector that can be well modeled by SVMs, resulting in significant gains in speaker recognition performance.

Original languageEnglish (US)
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume4
DOIs
StatePublished - 2007
Externally publishedYes
Event2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07 - Honolulu, HI, United States
Duration: Apr 15 2007Apr 20 2007

Other

Other2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07
CountryUnited States
CityHonolulu, HI
Period4/15/074/20/07

Fingerprint

Parameterization
parameterization
Support vector machines
syllables
evaluation

Keywords

  • GMM
  • Prosody
  • Speaker recognition
  • SVM

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Signal Processing
  • Acoustics and Ultrasonics

Cite this

Ferrer, L., Shriberg, E., Kajarekar, S., & Sonmez, M. K. (2007). Parameterization of prosodic feature distributions for SVM modeling in speaker recognition. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (Vol. 4). [4218080] https://doi.org/10.1109/ICASSP.2007.366892

Parameterization of prosodic feature distributions for SVM modeling in speaker recognition. / Ferrer, Luciana; Shriberg, Elizabeth; Kajarekar, Sachin; Sonmez, Mustafa (Kemal).

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 4 2007. 4218080.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ferrer, L, Shriberg, E, Kajarekar, S & Sonmez, MK 2007, Parameterization of prosodic feature distributions for SVM modeling in speaker recognition. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. vol. 4, 4218080, 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07, Honolulu, HI, United States, 4/15/07. https://doi.org/10.1109/ICASSP.2007.366892
Ferrer L, Shriberg E, Kajarekar S, Sonmez MK. Parameterization of prosodic feature distributions for SVM modeling in speaker recognition. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 4. 2007. 4218080 https://doi.org/10.1109/ICASSP.2007.366892
Ferrer, Luciana ; Shriberg, Elizabeth ; Kajarekar, Sachin ; Sonmez, Mustafa (Kemal). / Parameterization of prosodic feature distributions for SVM modeling in speaker recognition. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Vol. 4 2007.
@inproceedings{cb484941b53843e591e44fce89b06520,
title = "Parameterization of prosodic feature distributions for SVM modeling in speaker recognition",
abstract = "Multiple recent studies have shown that speaker recognition performance using frame-based cepstral features is improved by adding higher-level information, including prosodie and lexical features. This paper explores the important question of finding a good kernel for a system that models syllable-based prosodie features using support vector machines (SVMs). The system has been the best performing of our high-level systems in the last two NIST evaluations, and gives significant improvements when combined with cepstral-based systems. We introduce two new methods for transforming the syllable-level features into a single high-dimensional vector that can be well modeled by SVMs, resulting in significant gains in speaker recognition performance.",
keywords = "GMM, Prosody, Speaker recognition, SVM",
author = "Luciana Ferrer and Elizabeth Shriberg and Sachin Kajarekar and Sonmez, {Mustafa (Kemal)}",
year = "2007",
doi = "10.1109/ICASSP.2007.366892",
language = "English (US)",
isbn = "1424407281",
volume = "4",
booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

}

TY - GEN

T1 - Parameterization of prosodic feature distributions for SVM modeling in speaker recognition

AU - Ferrer, Luciana

AU - Shriberg, Elizabeth

AU - Kajarekar, Sachin

AU - Sonmez, Mustafa (Kemal)

PY - 2007

Y1 - 2007

N2 - Multiple recent studies have shown that speaker recognition performance using frame-based cepstral features is improved by adding higher-level information, including prosodie and lexical features. This paper explores the important question of finding a good kernel for a system that models syllable-based prosodie features using support vector machines (SVMs). The system has been the best performing of our high-level systems in the last two NIST evaluations, and gives significant improvements when combined with cepstral-based systems. We introduce two new methods for transforming the syllable-level features into a single high-dimensional vector that can be well modeled by SVMs, resulting in significant gains in speaker recognition performance.

AB - Multiple recent studies have shown that speaker recognition performance using frame-based cepstral features is improved by adding higher-level information, including prosodie and lexical features. This paper explores the important question of finding a good kernel for a system that models syllable-based prosodie features using support vector machines (SVMs). The system has been the best performing of our high-level systems in the last two NIST evaluations, and gives significant improvements when combined with cepstral-based systems. We introduce two new methods for transforming the syllable-level features into a single high-dimensional vector that can be well modeled by SVMs, resulting in significant gains in speaker recognition performance.

KW - GMM

KW - Prosody

KW - Speaker recognition

KW - SVM

UR - http://www.scopus.com/inward/record.url?scp=34547515912&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34547515912&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2007.366892

DO - 10.1109/ICASSP.2007.366892

M3 - Conference contribution

SN - 1424407281

SN - 9781424407286

VL - 4

BT - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

ER -