Using post-classifiers to enhance fusion of low- and high-level speaker recognition

Yosef A. Solewicz; Moshe Koppel

doi:10.1109/tasl.2007.903054

Using post-classifiers to enhance fusion of low- and high-level speaker recognition

Yosef A. Solewicz, Moshe Koppel

Department of Computer Science

Bar-Ilan University

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

7 Scopus citations

Abstract

This paper proposes a method for automatic correction of bias in speaker recognition systems, especially fusion-based systems. The method is based on a post-classifier which learns the relative performance obtained by the constituent systems in key trials, given the training and testing conditions in which they occurred. These conditions generally reflect train/test mismatch in factors such as channel, noise, speaker stress, etc. Results obtained with several state-of-the-art systems showed up to 20% decrease in EER compared to ordinary fusion in the NIST'05 Speaker Recognition Evaluation.

Original language	English
Article number	4291613
Pages (from-to)	2063-2071
Number of pages	9
Journal	IEEE Transactions on Audio, Speech and Language Processing
Volume	15
Issue number	7
DOIs	https://doi.org/10.1109/tasl.2007.903054
State	Published - Sep 2007

Keywords

Fusion
Machine learning
Post-classification
Speaker recognition

Access to Document

10.1109/tasl.2007.903054

Cite this

@article{be93be5b472f4b3f8ce9bdda50d304ba,

title = "Using post-classifiers to enhance fusion of low- and high-level speaker recognition",

abstract = "This paper proposes a method for automatic correction of bias in speaker recognition systems, especially fusion-based systems. The method is based on a post-classifier which learns the relative performance obtained by the constituent systems in key trials, given the training and testing conditions in which they occurred. These conditions generally reflect train/test mismatch in factors such as channel, noise, speaker stress, etc. Results obtained with several state-of-the-art systems showed up to 20% decrease in EER compared to ordinary fusion in the NIST'05 Speaker Recognition Evaluation.",

keywords = "Fusion, Machine learning, Post-classification, Speaker recognition",

author = "Solewicz, {Yosef A.} and Moshe Koppel",

year = "2007",

month = sep,

doi = "10.1109/tasl.2007.903054",

language = "אנגלית",

volume = "15",

pages = "2063--2071",

journal = "IEEE Transactions on Audio, Speech and Language Processing",

issn = "1558-7916",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "7",

}

TY - JOUR

T1 - Using post-classifiers to enhance fusion of low- and high-level speaker recognition

AU - Solewicz, Yosef A.

AU - Koppel, Moshe

PY - 2007/9

Y1 - 2007/9

N2 - This paper proposes a method for automatic correction of bias in speaker recognition systems, especially fusion-based systems. The method is based on a post-classifier which learns the relative performance obtained by the constituent systems in key trials, given the training and testing conditions in which they occurred. These conditions generally reflect train/test mismatch in factors such as channel, noise, speaker stress, etc. Results obtained with several state-of-the-art systems showed up to 20% decrease in EER compared to ordinary fusion in the NIST'05 Speaker Recognition Evaluation.

AB - This paper proposes a method for automatic correction of bias in speaker recognition systems, especially fusion-based systems. The method is based on a post-classifier which learns the relative performance obtained by the constituent systems in key trials, given the training and testing conditions in which they occurred. These conditions generally reflect train/test mismatch in factors such as channel, noise, speaker stress, etc. Results obtained with several state-of-the-art systems showed up to 20% decrease in EER compared to ordinary fusion in the NIST'05 Speaker Recognition Evaluation.

KW - Fusion

KW - Machine learning

KW - Post-classification

KW - Speaker recognition

UR - http://www.scopus.com/inward/record.url?scp=64249133832&partnerID=8YFLogxK

U2 - 10.1109/tasl.2007.903054

DO - 10.1109/tasl.2007.903054

M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???

AN - SCOPUS:64249133832

SN - 1558-7916

VL - 15

SP - 2063

EP - 2071

JO - IEEE Transactions on Audio, Speech and Language Processing

JF - IEEE Transactions on Audio, Speech and Language Processing

IS - 7

M1 - 4291613

ER -

Using post-classifiers to enhance fusion of low- and high-level speaker recognition

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this