Automatic speaker recognition for courtroom based on adaptive within-source-variance control

Hua Peng Wang; Jun Yang; Ming Wu; Yong Xu

doi:10.3969/j.issn.0255-8297.2014.06.006

Automatic speaker recognition for courtroom based on adaptive within-source-variance control

Hua Peng Wang^*, Jun Yang, Ming Wu, Yong Xu

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

Abstract

This paper proposes a method to transfer the scores generated from a speaker recognition system to likelihood ratios (LR) for evaluating the strength of forensic voice evidence. A robust LR estimation algorithm using adaptive within-source-variance control is developed to accurately estimate a model of the suspect. The algorithm adaptively combines information of reference speakers and that of the suspect to model the within source-variability of the suspect. Compared with a baseline recognition system, the system using the proposed algorithm has better discrimination capability and reliability, and the magnitude of evidence strength is also improved.

Original language	English
Pages (from-to)	582-587
Number of pages	6
Journal	Yingyong Kexue Xuebao/Journal of Applied Sciences
Volume	32
Issue number	6
DOIs	https://doi.org/10.3969/j.issn.0255-8297.2014.06.006
Publication status	Published - Nov 30 2014
Externally published	Yes

Bibliographical note

Publisher Copyright:
©, 2014, Press of Shanghai Scientific and Technical Publishers. All right reserved.

ASJC Scopus Subject Areas

General Computer Science
General Mathematics
General Engineering

Keywords

Adaptive within-source variance control
Background-model-Gaussian mixture model (BM-GMM)
Forensic automatic speaker recognition
Likelihood ratio

Access to Document

10.3969/j.issn.0255-8297.2014.06.006

Cite this

@article{2ff6529465b44288a3a12251408da87c,

title = "Automatic speaker recognition for courtroom based on adaptive within-source-variance control",

abstract = "This paper proposes a method to transfer the scores generated from a speaker recognition system to likelihood ratios (LR) for evaluating the strength of forensic voice evidence. A robust LR estimation algorithm using adaptive within-source-variance control is developed to accurately estimate a model of the suspect. The algorithm adaptively combines information of reference speakers and that of the suspect to model the within source-variability of the suspect. Compared with a baseline recognition system, the system using the proposed algorithm has better discrimination capability and reliability, and the magnitude of evidence strength is also improved.",

keywords = "Adaptive within-source variance control, Background-model-Gaussian mixture model (BM-GMM), Forensic automatic speaker recognition, Likelihood ratio",

author = "Wang, \{Hua Peng\} and Jun Yang and Ming Wu and Yong Xu",

year = "2014",

month = nov,

day = "30",

doi = "10.3969/j.issn.0255-8297.2014.06.006",

language = "English",

volume = "32",

pages = "582--587",

journal = "Yingyong Kexue Xuebao/Journal of Applied Sciences",

issn = "0255-8297",

publisher = "Shanghai Science and Technology Press",

number = "6",

}

TY - JOUR

T1 - Automatic speaker recognition for courtroom based on adaptive within-source-variance control

AU - Wang, Hua Peng

AU - Yang, Jun

AU - Wu, Ming

AU - Xu, Yong

PY - 2014/11/30

Y1 - 2014/11/30

N2 - This paper proposes a method to transfer the scores generated from a speaker recognition system to likelihood ratios (LR) for evaluating the strength of forensic voice evidence. A robust LR estimation algorithm using adaptive within-source-variance control is developed to accurately estimate a model of the suspect. The algorithm adaptively combines information of reference speakers and that of the suspect to model the within source-variability of the suspect. Compared with a baseline recognition system, the system using the proposed algorithm has better discrimination capability and reliability, and the magnitude of evidence strength is also improved.

AB - This paper proposes a method to transfer the scores generated from a speaker recognition system to likelihood ratios (LR) for evaluating the strength of forensic voice evidence. A robust LR estimation algorithm using adaptive within-source-variance control is developed to accurately estimate a model of the suspect. The algorithm adaptively combines information of reference speakers and that of the suspect to model the within source-variability of the suspect. Compared with a baseline recognition system, the system using the proposed algorithm has better discrimination capability and reliability, and the magnitude of evidence strength is also improved.

KW - Adaptive within-source variance control

KW - Background-model-Gaussian mixture model (BM-GMM)

KW - Forensic automatic speaker recognition

KW - Likelihood ratio

UR - http://www.scopus.com/inward/record.url?scp=84916597617&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84916597617&partnerID=8YFLogxK

U2 - 10.3969/j.issn.0255-8297.2014.06.006

DO - 10.3969/j.issn.0255-8297.2014.06.006

M3 - Article

AN - SCOPUS:84916597617

SN - 0255-8297

VL - 32

SP - 582

EP - 587

JO - Yingyong Kexue Xuebao/Journal of Applied Sciences

JF - Yingyong Kexue Xuebao/Journal of Applied Sciences

IS - 6

ER -

Automatic speaker recognition for courtroom based on adaptive within-source-variance control

Abstract

Bibliographical note

ASJC Scopus Subject Areas

Keywords

Access to Document

Other files and links

Fingerprint

Cite this