NEC-TT speaker verification system for SRE'19 CTS challenge

Kong Aik Lee; Koji Okabe; Hitoshi Yamamoto; Qiongqiong Wang; Ling Guo; Takafumi Koshinaka; Jiacen Zhang; Keisuke Ishikawa; Koichi Shinoda

doi:10.21437/Interspeech.2020-1132

NEC-TT speaker verification system for SRE'19 CTS challenge

Kong Aik Lee, Koji Okabe, Hitoshi Yamamoto, Qiongqiong Wang, Ling Guo, Takafumi Koshinaka, Jiacen Zhang, Keisuke Ishikawa, Koichi Shinoda

Research output: Journal article publication › Conference article › Academic research › peer-review

5 Citations (Scopus)

Abstract

The series of speaker recognition evaluations (SREs) organized by the National Institute of Standards and Technology (NIST) is widely accepted as the de facto benchmark for speaker recognition technology. This paper describes the NEC-TT speaker verification system developed for the recent SRE'19 CTS Challenge. Our system is based on an x-vector embedding front-end followed by a thin scoring back-end. We trained a very-deep neural network for x-vector extraction by incorporating residual connections, squeeze-and-excitation networks, and angular-margin softmax at the output layer. We enhanced the back-end with a tandem approach leveraging the benefit of supervised and unsupervised domain adaptation. We obtained over 30% relative reduction in error rate with each of these enhancements at the front-end and back-end, respectively.

Original language	English
Pages (from-to)	2227-2231
Number of pages	5
Journal	Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
Volume	2020-October
DOIs	https://doi.org/10.21437/Interspeech.2020-1132
Publication status	Published - Oct 2020
Externally published	Yes
Event	21st Annual Conference of the International Speech Communication Association, INTERSPEECH 2020 - Shanghai, China Duration: 25 Oct 2020 → 29 Oct 2020

Keywords

Benchmark evaluation
Speaker recognition

ASJC Scopus subject areas

Language and Linguistics
Human-Computer Interaction
Signal Processing
Software
Modelling and Simulation

More information

10.21437/Interspeech.2020-1132

Cite this

@article{af382c31dbb346efabeea8fe83d59b07,

title = "NEC-TT speaker verification system for SRE'19 CTS challenge",

abstract = "The series of speaker recognition evaluations (SREs) organized by the National Institute of Standards and Technology (NIST) is widely accepted as the de facto benchmark for speaker recognition technology. This paper describes the NEC-TT speaker verification system developed for the recent SRE'19 CTS Challenge. Our system is based on an x-vector embedding front-end followed by a thin scoring back-end. We trained a very-deep neural network for x-vector extraction by incorporating residual connections, squeeze-and-excitation networks, and angular-margin softmax at the output layer. We enhanced the back-end with a tandem approach leveraging the benefit of supervised and unsupervised domain adaptation. We obtained over 30% relative reduction in error rate with each of these enhancements at the front-end and back-end, respectively.",

keywords = "Benchmark evaluation, Speaker recognition",

author = "Lee, {Kong Aik} and Koji Okabe and Hitoshi Yamamoto and Qiongqiong Wang and Ling Guo and Takafumi Koshinaka and Jiacen Zhang and Keisuke Ishikawa and Koichi Shinoda",

note = "Publisher Copyright: Copyright {\textcopyright} 2020 ISCA; 21st Annual Conference of the International Speech Communication Association, INTERSPEECH 2020 ; Conference date: 25-10-2020 Through 29-10-2020",

year = "2020",

month = oct,

doi = "10.21437/Interspeech.2020-1132",

language = "English",

volume = "2020-October",

pages = "2227--2231",

journal = "Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH",

issn = "2308-457X",

}

NEC-TT speaker verification system for SRE'19 CTS challenge. / Lee, Kong Aik; Okabe, Koji; Yamamoto, Hitoshi et al.
In: Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, Vol. 2020-October, 10.2020, p. 2227-2231.

Research output: Journal article publication › Conference article › Academic research › peer-review

TY - JOUR

T1 - NEC-TT speaker verification system for SRE'19 CTS challenge

AU - Lee, Kong Aik

AU - Okabe, Koji

AU - Yamamoto, Hitoshi

AU - Wang, Qiongqiong

AU - Guo, Ling

AU - Koshinaka, Takafumi

AU - Zhang, Jiacen

AU - Ishikawa, Keisuke

AU - Shinoda, Koichi

PY - 2020/10

Y1 - 2020/10

N2 - The series of speaker recognition evaluations (SREs) organized by the National Institute of Standards and Technology (NIST) is widely accepted as the de facto benchmark for speaker recognition technology. This paper describes the NEC-TT speaker verification system developed for the recent SRE'19 CTS Challenge. Our system is based on an x-vector embedding front-end followed by a thin scoring back-end. We trained a very-deep neural network for x-vector extraction by incorporating residual connections, squeeze-and-excitation networks, and angular-margin softmax at the output layer. We enhanced the back-end with a tandem approach leveraging the benefit of supervised and unsupervised domain adaptation. We obtained over 30% relative reduction in error rate with each of these enhancements at the front-end and back-end, respectively.

AB - The series of speaker recognition evaluations (SREs) organized by the National Institute of Standards and Technology (NIST) is widely accepted as the de facto benchmark for speaker recognition technology. This paper describes the NEC-TT speaker verification system developed for the recent SRE'19 CTS Challenge. Our system is based on an x-vector embedding front-end followed by a thin scoring back-end. We trained a very-deep neural network for x-vector extraction by incorporating residual connections, squeeze-and-excitation networks, and angular-margin softmax at the output layer. We enhanced the back-end with a tandem approach leveraging the benefit of supervised and unsupervised domain adaptation. We obtained over 30% relative reduction in error rate with each of these enhancements at the front-end and back-end, respectively.

KW - Benchmark evaluation

KW - Speaker recognition

UR - http://www.scopus.com/inward/record.url?scp=85098116717&partnerID=8YFLogxK

U2 - 10.21437/Interspeech.2020-1132

DO - 10.21437/Interspeech.2020-1132

M3 - Conference article

AN - SCOPUS:85098116717

SN - 2308-457X

VL - 2020-October

SP - 2227

EP - 2231

JO - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH

JF - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH

T2 - 21st Annual Conference of the International Speech Communication Association, INTERSPEECH 2020

Y2 - 25 October 2020 through 29 October 2020

ER -

NEC-TT speaker verification system for SRE'19 CTS challenge

Abstract

Keywords

ASJC Scopus subject areas

More information

Other files and links

Fingerprint

Cite this