Nonlinear kernel nuisance attribute projection for speaker verification

Xianyu Zhao; Yuan Dong; Hao Yang; Jian Zhao; Liang Lu; Haila Wang

doi:10.1109/ICASSP.2008.4518562

Nonlinear kernel nuisance attribute projection for speaker verification

Xianyu Zhao, Yuan Dong, Hao Yang, Jian Zhao, Liang Lu, Haila Wang

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Nuisance attribute projection (NAP) was successfully applied in SVM-based speaker verification systems to improve performance by doing projection to remove dimensions from the SVM feature space that cause unwanted variability in the kernel. Previous studies of NAP were focused mainly on linear and generalized linear kernel SVMs. In this paper, NAP in nonlinear kernel SVMs, e.g. polynomial or Gaussian kernels, are investigated. Instead of doing explicit feature expansion and projection in high-dimension feature space, kernel principal component analysis is employed to find nuisance dimensions; and, NAP is carried out implicitly by incorporating it into some compensated kernel functions. Experimental results on the 2006 NIST SRE corpus indicate the effectiveness of such nonlinear kernel NAP. Compared with linear NAP, nonlinear NAP with Gaussian kernel obtained about 11% relative improvement in equal error rate (EER).

Original language	English
Title of host publication	2008 IEEE International Conference on Acoustics, Speech and Signal Processing
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Pages	4125-4128
Number of pages	4
ISBN (Electronic)	978-1-4244-1484-0
ISBN (Print)	978-1-4244-1483-3
DOIs	https://doi.org/10.1109/ICASSP.2008.4518562
Publication status	Published - 1 Mar 2008
Event	2008 IEEE International Conference on Acoustics, Speech and Signal Processing - Caesars Palace, Las Vegas, NV, United States Duration: 30 Mar 2008 → 4 Apr 2008

Conference

Conference	2008 IEEE International Conference on Acoustics, Speech and Signal Processing
Country	United States
City	Las Vegas, NV
Period	30/03/08 → 4/04/08

Keywords

Gaussian processes
polynomials
principal component analysis
speaker recognition
Gaussian kernels
SVM-based speaker verification systems
equal error rate
high-dimension feature space
kernel principal component analysis
nonlinear kernel
nonlinear kernel nuisance attribute projection
polynomial kernels
speaker verification
Input variables
Kernel
NIST
Polynomials
Principal component analysis
Research and development
Speaker recognition
Support vector machine classification
Support vector machines
Telecommunications
nuisance attribute projection
speaker location
supporting vector machines

Access to Document

10.1109/ICASSP.2008.4518562

Fingerprint Dive into the research topics of 'Nonlinear kernel nuisance attribute projection for speaker verification'. Together they form a unique fingerprint.

Cite this

@inproceedings{ed2c549d3a7d45a9abf4e29c568c8b32,

title = "Nonlinear kernel nuisance attribute projection for speaker verification",

abstract = "Nuisance attribute projection (NAP) was successfully applied in SVM-based speaker verification systems to improve performance by doing projection to remove dimensions from the SVM feature space that cause unwanted variability in the kernel. Previous studies of NAP were focused mainly on linear and generalized linear kernel SVMs. In this paper, NAP in nonlinear kernel SVMs, e.g. polynomial or Gaussian kernels, are investigated. Instead of doing explicit feature expansion and projection in high-dimension feature space, kernel principal component analysis is employed to find nuisance dimensions; and, NAP is carried out implicitly by incorporating it into some compensated kernel functions. Experimental results on the 2006 NIST SRE corpus indicate the effectiveness of such nonlinear kernel NAP. Compared with linear NAP, nonlinear NAP with Gaussian kernel obtained about 11% relative improvement in equal error rate (EER).",

keywords = "Gaussian processes, polynomials, principal component analysis, speaker recognition, Gaussian kernels, SVM-based speaker verification systems, equal error rate, high-dimension feature space, kernel principal component analysis, nonlinear kernel, nonlinear kernel nuisance attribute projection, polynomial kernels, speaker verification, Input variables, Kernel, NIST, Polynomials, Principal component analysis, Research and development, Speaker recognition, Support vector machine classification, Support vector machines, Telecommunications, nuisance attribute projection, speaker location, supporting vector machines",

author = "Xianyu Zhao and Yuan Dong and Hao Yang and Jian Zhao and Liang Lu and Haila Wang",

year = "2008",

month = mar,

day = "1",

doi = "10.1109/ICASSP.2008.4518562",

language = "English",

isbn = "978-1-4244-1483-3",

pages = "4125--4128",

booktitle = "2008 IEEE International Conference on Acoustics, Speech and Signal Processing",

publisher = "Institute of Electrical and Electronics Engineers (IEEE)",

address = "United States",

note = "2008 IEEE International Conference on Acoustics, Speech and Signal Processing ; Conference date: 30-03-2008 Through 04-04-2008",

}

Zhao, X, Dong, Y, Yang, H, Zhao, J, Lu, L & Wang, H 2008, Nonlinear kernel nuisance attribute projection for speaker verification. in 2008 IEEE International Conference on Acoustics, Speech and Signal Processing. Institute of Electrical and Electronics Engineers (IEEE), pp. 4125-4128, 2008 IEEE International Conference on Acoustics, Speech and Signal Processing, Las Vegas, NV, United States, 30/03/08. https://doi.org/10.1109/ICASSP.2008.4518562

Nonlinear kernel nuisance attribute projection for speaker verification. / Zhao, Xianyu; Dong, Yuan; Yang, Hao; Zhao, Jian; Lu, Liang; Wang, Haila.

2008 IEEE International Conference on Acoustics, Speech and Signal Processing. Institute of Electrical and Electronics Engineers (IEEE), 2008. p. 4125-4128.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Nonlinear kernel nuisance attribute projection for speaker verification

AU - Zhao, Xianyu

AU - Dong, Yuan

AU - Yang, Hao

AU - Zhao, Jian

AU - Lu, Liang

AU - Wang, Haila

PY - 2008/3/1

Y1 - 2008/3/1

N2 - Nuisance attribute projection (NAP) was successfully applied in SVM-based speaker verification systems to improve performance by doing projection to remove dimensions from the SVM feature space that cause unwanted variability in the kernel. Previous studies of NAP were focused mainly on linear and generalized linear kernel SVMs. In this paper, NAP in nonlinear kernel SVMs, e.g. polynomial or Gaussian kernels, are investigated. Instead of doing explicit feature expansion and projection in high-dimension feature space, kernel principal component analysis is employed to find nuisance dimensions; and, NAP is carried out implicitly by incorporating it into some compensated kernel functions. Experimental results on the 2006 NIST SRE corpus indicate the effectiveness of such nonlinear kernel NAP. Compared with linear NAP, nonlinear NAP with Gaussian kernel obtained about 11% relative improvement in equal error rate (EER).

AB - Nuisance attribute projection (NAP) was successfully applied in SVM-based speaker verification systems to improve performance by doing projection to remove dimensions from the SVM feature space that cause unwanted variability in the kernel. Previous studies of NAP were focused mainly on linear and generalized linear kernel SVMs. In this paper, NAP in nonlinear kernel SVMs, e.g. polynomial or Gaussian kernels, are investigated. Instead of doing explicit feature expansion and projection in high-dimension feature space, kernel principal component analysis is employed to find nuisance dimensions; and, NAP is carried out implicitly by incorporating it into some compensated kernel functions. Experimental results on the 2006 NIST SRE corpus indicate the effectiveness of such nonlinear kernel NAP. Compared with linear NAP, nonlinear NAP with Gaussian kernel obtained about 11% relative improvement in equal error rate (EER).

KW - Gaussian processes

KW - polynomials

KW - principal component analysis

KW - speaker recognition

KW - Gaussian kernels

KW - SVM-based speaker verification systems

KW - equal error rate

KW - high-dimension feature space

KW - kernel principal component analysis

KW - nonlinear kernel

KW - nonlinear kernel nuisance attribute projection

KW - polynomial kernels

KW - speaker verification

KW - Input variables

KW - Kernel

KW - NIST

KW - Polynomials

KW - Principal component analysis

KW - Research and development

KW - Speaker recognition

KW - Support vector machine classification

KW - Support vector machines

KW - Telecommunications

KW - nuisance attribute projection

KW - speaker location

KW - supporting vector machines

U2 - 10.1109/ICASSP.2008.4518562

DO - 10.1109/ICASSP.2008.4518562

M3 - Conference contribution

SN - 978-1-4244-1483-3

SP - 4125

EP - 4128

BT - 2008 IEEE International Conference on Acoustics, Speech and Signal Processing

PB - Institute of Electrical and Electronics Engineers (IEEE)

T2 - 2008 IEEE International Conference on Acoustics, Speech and Signal Processing

Y2 - 30 March 2008 through 4 April 2008

ER -