Speaker verification with TIMIT corpus-some remarks on classical methods

Adam Dustor

doi:10.23919/spa50552.2020.9241298

Speaker verification with TIMIT corpus-some remarks on classical methods

Adam Dustor

Department of Telecommunications and Teleinformatics (RAU/RAU13)

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

2 Citations (Scopus)

Abstract

The aim of this paper is to present some research on speaker verification system based on Gaussian Mixture Model-Universal Background Model (GMM-UBM) approach. All tests were done for the TIMIT corpus. Performance for the standard Mel-Frequency Cepstral Coefficients (MFCC) and dynamic delta features is shown. Influence of feature dimensionality and model complexity on Equal Error Rate (EER) is presented. Additionally, an impact of Voice Activity Detection (VAD) and normalization techniques like Cepstral Mean and Variance Normalization (CMVN) and RelAtive SpecTrA (RASTA) filtering is covered. Each combination of factors was examined. It is shown that careful selection of traditional techniques may lead to very satisfying results when it comes to achieved EER values.

Original language	English
Title of host publication	SPA 2020 - Signal Processing
Subtitle of host publication	Algorithms, Architectures, Arrangements, and Applications, Conference Proceedings
Publisher	IEEE Computer Society
Pages	174-179
Number of pages	6
ISBN (Electronic)	9788362065394
DOIs	https://doi.org/10.23919/spa50552.2020.9241298
Publication status	Published - 23 Sept 2020
Event	24th IEEE Signal Processing: Algorithms, Architectures, Arrangements, and Applications, SPA 2020 - Virtual, Poznan, Poland Duration: 23 Sept 2020 → 25 Sept 2020

Publication series

Name	Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA
Volume	2020-September
ISSN (Print)	2326-0262
ISSN (Electronic)	2326-0319

Conference

Conference	24th IEEE Signal Processing: Algorithms, Architectures, Arrangements, and Applications, SPA 2020
Country/Territory	Poland
City	Virtual, Poznan
Period	23/09/20 → 25/09/20

ASJC Scopus subject areas

Computational Theory and Mathematics
Computer Science Applications
Information Systems
Signal Processing

Access to Document

10.23919/spa50552.2020.9241298

Cite this

Dustor, A. (2020). Speaker verification with TIMIT corpus-some remarks on classical methods. In SPA 2020 - Signal Processing: Algorithms, Architectures, Arrangements, and Applications, Conference Proceedings (pp. 174-179). Article 9241298 (Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA; Vol. 2020-September). IEEE Computer Society. https://doi.org/10.23919/spa50552.2020.9241298

@inproceedings{a5731d30eb234f0c8c718f32ca28765a,

title = "Speaker verification with TIMIT corpus-some remarks on classical methods",

abstract = "The aim of this paper is to present some research on speaker verification system based on Gaussian Mixture Model-Universal Background Model (GMM-UBM) approach. All tests were done for the TIMIT corpus. Performance for the standard Mel-Frequency Cepstral Coefficients (MFCC) and dynamic delta features is shown. Influence of feature dimensionality and model complexity on Equal Error Rate (EER) is presented. Additionally, an impact of Voice Activity Detection (VAD) and normalization techniques like Cepstral Mean and Variance Normalization (CMVN) and RelAtive SpecTrA (RASTA) filtering is covered. Each combination of factors was examined. It is shown that careful selection of traditional techniques may lead to very satisfying results when it comes to achieved EER values.",

author = "Adam Dustor",

note = "Publisher Copyright: {\textcopyright} 2020 Division of Signal Processing and Electronic Systems, Poznan University of Technology (DSPES PUT).; 24th IEEE Signal Processing: Algorithms, Architectures, Arrangements, and Applications, SPA 2020 ; Conference date: 23-09-2020 Through 25-09-2020",

year = "2020",

month = sep,

day = "23",

doi = "10.23919/spa50552.2020.9241298",

language = "English",

series = "Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA",

publisher = "IEEE Computer Society",

pages = "174--179",

booktitle = "SPA 2020 - Signal Processing",

address = "United States",

}

Dustor, A 2020, Speaker verification with TIMIT corpus-some remarks on classical methods. in SPA 2020 - Signal Processing: Algorithms, Architectures, Arrangements, and Applications, Conference Proceedings., 9241298, Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA, vol. 2020-September, IEEE Computer Society, pp. 174-179, 24th IEEE Signal Processing: Algorithms, Architectures, Arrangements, and Applications, SPA 2020, Virtual, Poznan, Poland, 23/09/20. https://doi.org/10.23919/spa50552.2020.9241298

Speaker verification with TIMIT corpus-some remarks on classical methods. / Dustor, Adam.
SPA 2020 - Signal Processing: Algorithms, Architectures, Arrangements, and Applications, Conference Proceedings. IEEE Computer Society, 2020. p. 174-179 9241298 (Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA; Vol. 2020-September).

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

TY - GEN

T1 - Speaker verification with TIMIT corpus-some remarks on classical methods

AU - Dustor, Adam

PY - 2020/9/23

Y1 - 2020/9/23

N2 - The aim of this paper is to present some research on speaker verification system based on Gaussian Mixture Model-Universal Background Model (GMM-UBM) approach. All tests were done for the TIMIT corpus. Performance for the standard Mel-Frequency Cepstral Coefficients (MFCC) and dynamic delta features is shown. Influence of feature dimensionality and model complexity on Equal Error Rate (EER) is presented. Additionally, an impact of Voice Activity Detection (VAD) and normalization techniques like Cepstral Mean and Variance Normalization (CMVN) and RelAtive SpecTrA (RASTA) filtering is covered. Each combination of factors was examined. It is shown that careful selection of traditional techniques may lead to very satisfying results when it comes to achieved EER values.

AB - The aim of this paper is to present some research on speaker verification system based on Gaussian Mixture Model-Universal Background Model (GMM-UBM) approach. All tests were done for the TIMIT corpus. Performance for the standard Mel-Frequency Cepstral Coefficients (MFCC) and dynamic delta features is shown. Influence of feature dimensionality and model complexity on Equal Error Rate (EER) is presented. Additionally, an impact of Voice Activity Detection (VAD) and normalization techniques like Cepstral Mean and Variance Normalization (CMVN) and RelAtive SpecTrA (RASTA) filtering is covered. Each combination of factors was examined. It is shown that careful selection of traditional techniques may lead to very satisfying results when it comes to achieved EER values.

UR - https://www.scopus.com/pages/publications/85096242475

U2 - 10.23919/spa50552.2020.9241298

DO - 10.23919/spa50552.2020.9241298

M3 - Conference contribution

AN - SCOPUS:85096242475

T3 - Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA

SP - 174

EP - 179

BT - SPA 2020 - Signal Processing

PB - IEEE Computer Society

T2 - 24th IEEE Signal Processing: Algorithms, Architectures, Arrangements, and Applications, SPA 2020

Y2 - 23 September 2020 through 25 September 2020

ER -

Dustor A. Speaker verification with TIMIT corpus-some remarks on classical methods. In SPA 2020 - Signal Processing: Algorithms, Architectures, Arrangements, and Applications, Conference Proceedings. IEEE Computer Society. 2020. p. 174-179. 9241298. (Signal Processing - Algorithms, Architectures, Arrangements, and Applications Conference Proceedings, SPA). doi: 10.23919/spa50552.2020.9241298

Speaker verification with TIMIT corpus-some remarks on classical methods

Abstract

Publication series

Conference

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this