Robust Session Variability Compensation for SVM Speaker Verification

Hyunson Seo, Chi Sang Jung, Hong-Goo Kang

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

This paper presents an enhanced nuisance attribute projection (NAP) method to improve the performance of speaker verification systems in mismatched train and test conditions. Unlike the conventional NAP training method that does not take any scheme to discriminate the source of nuisance, the proposed method quantitatively estimates the source of nuisance based on the statistics of given background speakers' eigenvalues. The estimated values are used for defining a discriminative weight for each of background speakers and selectively including the statistics of between-class scatter or of within-class scatter from them. Through the scheme, we intend to design a more robust projection matrix which involves less speaker-dependent or speaker-intrinsic variability while including more latent nuisance factors beyond the common within-class scatter of backgrounds. Experimental results on the recent NIST SRE evaluations demonstrate that the proposed algorithms produce consistent improvement over the previous NAP approaches.

Original languageEnglish
Pages (from-to)1631-1641
Number of pages11
JournalIEEE Transactions on Audio, Speech and Language Processing
Volume19
Issue number6
DOIs
Publication statusPublished - 2011 Jan 1

Fingerprint

projection
Statistics
statistics
education
eigenvalues
Compensation and Redress
evaluation
estimates
matrices

All Science Journal Classification (ASJC) codes

  • Acoustics and Ultrasonics
  • Electrical and Electronic Engineering

Cite this

@article{85c3703820284149aa7297e5f99c4ea9,
title = "Robust Session Variability Compensation for SVM Speaker Verification",
abstract = "This paper presents an enhanced nuisance attribute projection (NAP) method to improve the performance of speaker verification systems in mismatched train and test conditions. Unlike the conventional NAP training method that does not take any scheme to discriminate the source of nuisance, the proposed method quantitatively estimates the source of nuisance based on the statistics of given background speakers' eigenvalues. The estimated values are used for defining a discriminative weight for each of background speakers and selectively including the statistics of between-class scatter or of within-class scatter from them. Through the scheme, we intend to design a more robust projection matrix which involves less speaker-dependent or speaker-intrinsic variability while including more latent nuisance factors beyond the common within-class scatter of backgrounds. Experimental results on the recent NIST SRE evaluations demonstrate that the proposed algorithms produce consistent improvement over the previous NAP approaches.",
author = "Hyunson Seo and Jung, {Chi Sang} and Hong-Goo Kang",
year = "2011",
month = "1",
day = "1",
doi = "10.1109/TASL.2010.2093895",
language = "English",
volume = "19",
pages = "1631--1641",
journal = "IEEE Transactions on Speech and Audio Processing",
issn = "1558-7916",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "6",

}

Robust Session Variability Compensation for SVM Speaker Verification. / Seo, Hyunson; Jung, Chi Sang; Kang, Hong-Goo.

In: IEEE Transactions on Audio, Speech and Language Processing, Vol. 19, No. 6, 01.01.2011, p. 1631-1641.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Robust Session Variability Compensation for SVM Speaker Verification

AU - Seo, Hyunson

AU - Jung, Chi Sang

AU - Kang, Hong-Goo

PY - 2011/1/1

Y1 - 2011/1/1

N2 - This paper presents an enhanced nuisance attribute projection (NAP) method to improve the performance of speaker verification systems in mismatched train and test conditions. Unlike the conventional NAP training method that does not take any scheme to discriminate the source of nuisance, the proposed method quantitatively estimates the source of nuisance based on the statistics of given background speakers' eigenvalues. The estimated values are used for defining a discriminative weight for each of background speakers and selectively including the statistics of between-class scatter or of within-class scatter from them. Through the scheme, we intend to design a more robust projection matrix which involves less speaker-dependent or speaker-intrinsic variability while including more latent nuisance factors beyond the common within-class scatter of backgrounds. Experimental results on the recent NIST SRE evaluations demonstrate that the proposed algorithms produce consistent improvement over the previous NAP approaches.

AB - This paper presents an enhanced nuisance attribute projection (NAP) method to improve the performance of speaker verification systems in mismatched train and test conditions. Unlike the conventional NAP training method that does not take any scheme to discriminate the source of nuisance, the proposed method quantitatively estimates the source of nuisance based on the statistics of given background speakers' eigenvalues. The estimated values are used for defining a discriminative weight for each of background speakers and selectively including the statistics of between-class scatter or of within-class scatter from them. Through the scheme, we intend to design a more robust projection matrix which involves less speaker-dependent or speaker-intrinsic variability while including more latent nuisance factors beyond the common within-class scatter of backgrounds. Experimental results on the recent NIST SRE evaluations demonstrate that the proposed algorithms produce consistent improvement over the previous NAP approaches.

UR - http://www.scopus.com/inward/record.url?scp=85008025314&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85008025314&partnerID=8YFLogxK

U2 - 10.1109/TASL.2010.2093895

DO - 10.1109/TASL.2010.2093895

M3 - Article

AN - SCOPUS:85008025314

VL - 19

SP - 1631

EP - 1641

JO - IEEE Transactions on Speech and Audio Processing

JF - IEEE Transactions on Speech and Audio Processing

SN - 1558-7916

IS - 6

ER -