Speaker-dependent Dictionary-based Speech Enhancement for Text-Dependent Speaker Verification

Nicolai Bæk Thomsen, Dennis Alexander Lehmann Thomsen, Zheng-Hua Tan, Børge Lindberg, Søren Holdt Jensen

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

4 Citations (Scopus)

Abstract

The problem of text-dependent speaker verification under noisy conditions is becoming ever more relevant, due to increased
usage for authentication in real-world applications.
Classical methods for noise reduction such as spectral subtraction and Wiener filtering introduce distortion and do not perform well in
this setting.
In this work we compare the performance of different noise reduction methods under different noise conditions
in terms of speaker verification when the text is known and the system is trained on clean data (mis-matched conditions).
We furthermore propose a new approach based on dictionary-based noise reduction and compare it to the baseline methods.
Original languageEnglish
Title of host publicationInterspeech 2016
PublisherISCA
Publication dateSep 2016
Pages1839-1843
DOIs
Publication statusPublished - Sep 2016
EventInterspeech 2016 - San Francisco, CA, United States
Duration: 8 Sep 201612 Sep 2016
http://www.interspeech2016.org/

Conference

ConferenceInterspeech 2016
CountryUnited States
CitySan Francisco, CA
Period08/09/201612/09/2016
Internet address

Fingerprint

Speech enhancement
Glossaries
Noise abatement
Acoustic noise
Authentication

Keywords

  • Noise reduction
  • speaker verification
  • dictionary learning

Cite this

Thomsen, Nicolai Bæk ; Thomsen, Dennis Alexander Lehmann ; Tan, Zheng-Hua ; Lindberg, Børge ; Jensen, Søren Holdt. / Speaker-dependent Dictionary-based Speech Enhancement for Text-Dependent Speaker Verification. Interspeech 2016. ISCA, 2016. pp. 1839-1843
@inproceedings{7de9d9ebf2bc47e1b8e6a3590282efb2,
title = "Speaker-dependent Dictionary-based Speech Enhancement for Text-Dependent Speaker Verification",
abstract = "The problem of text-dependent speaker verification under noisy conditions is becoming ever more relevant, due to increasedusage for authentication in real-world applications. Classical methods for noise reduction such as spectral subtraction and Wiener filtering introduce distortion and do not perform well inthis setting.In this work we compare the performance of different noise reduction methods under different noise conditionsin terms of speaker verification when the text is known and the system is trained on clean data (mis-matched conditions).We furthermore propose a new approach based on dictionary-based noise reduction and compare it to the baseline methods.",
keywords = "Noise reduction, speaker verification, dictionary learning",
author = "Thomsen, {Nicolai B{\ae}k} and Thomsen, {Dennis Alexander Lehmann} and Zheng-Hua Tan and B{\o}rge Lindberg and Jensen, {S{\o}ren Holdt}",
year = "2016",
month = "9",
doi = "10.21437/Interspeech.2016-763",
language = "English",
pages = "1839--1843",
booktitle = "Interspeech 2016",
publisher = "ISCA",

}

Thomsen, NB, Thomsen, DAL, Tan, Z-H, Lindberg, B & Jensen, SH 2016, Speaker-dependent Dictionary-based Speech Enhancement for Text-Dependent Speaker Verification. in Interspeech 2016. ISCA, pp. 1839-1843, Interspeech 2016, San Francisco, CA, United States, 08/09/2016. https://doi.org/10.21437/Interspeech.2016-763

Speaker-dependent Dictionary-based Speech Enhancement for Text-Dependent Speaker Verification. / Thomsen, Nicolai Bæk; Thomsen, Dennis Alexander Lehmann; Tan, Zheng-Hua; Lindberg, Børge; Jensen, Søren Holdt.

Interspeech 2016. ISCA, 2016. p. 1839-1843.

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - Speaker-dependent Dictionary-based Speech Enhancement for Text-Dependent Speaker Verification

AU - Thomsen, Nicolai Bæk

AU - Thomsen, Dennis Alexander Lehmann

AU - Tan, Zheng-Hua

AU - Lindberg, Børge

AU - Jensen, Søren Holdt

PY - 2016/9

Y1 - 2016/9

N2 - The problem of text-dependent speaker verification under noisy conditions is becoming ever more relevant, due to increasedusage for authentication in real-world applications. Classical methods for noise reduction such as spectral subtraction and Wiener filtering introduce distortion and do not perform well inthis setting.In this work we compare the performance of different noise reduction methods under different noise conditionsin terms of speaker verification when the text is known and the system is trained on clean data (mis-matched conditions).We furthermore propose a new approach based on dictionary-based noise reduction and compare it to the baseline methods.

AB - The problem of text-dependent speaker verification under noisy conditions is becoming ever more relevant, due to increasedusage for authentication in real-world applications. Classical methods for noise reduction such as spectral subtraction and Wiener filtering introduce distortion and do not perform well inthis setting.In this work we compare the performance of different noise reduction methods under different noise conditionsin terms of speaker verification when the text is known and the system is trained on clean data (mis-matched conditions).We furthermore propose a new approach based on dictionary-based noise reduction and compare it to the baseline methods.

KW - Noise reduction

KW - speaker verification

KW - dictionary learning

U2 - 10.21437/Interspeech.2016-763

DO - 10.21437/Interspeech.2016-763

M3 - Article in proceeding

SP - 1839

EP - 1843

BT - Interspeech 2016

PB - ISCA

ER -