UIAI System for Short-Duration Speaker Verification Challenge 2020

Md Sahidullah; Achintya Kumar Sarkar; Ville Vestman; Xuechen Liu; Romain Serizel; Tomi Kinnunen; Zheng-Hua Tan; Emmanuel Vincent

doi:10.1109/SLT48900.2021.9383596

UIAI System for Short-Duration Speaker Verification Challenge 2020

Md Sahidullah, Achintya Kumar Sarkar, Ville Vestman, Xuechen Liu, Romain Serizel, Tomi Kinnunen, Zheng-Hua Tan, Emmanuel Vincent

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

6 Citations (Scopus)

Abstract

In this work, we present the system description of the UIAI entry for the short-duration speaker verification (SdSV) challenge 2020. Our focus is on Task 1 dedicated to text-dependent speaker verification. We investigate different feature extraction and modeling approaches for automatic speaker verification (ASV) and utterance verification (UV). We have also studied different fusion strategies for combining UV and ASV modules. Our primary submission to the challenge is the fusion of seven subsystems which yields a normalized minimum detection cost function (minDCF) of 0.072 and an equal error rate (EER) of 2.14% on the evaluation set. The single system consisting of a pass-phrase identification based model with phone-discriminative bottleneck features gives a normalized minDCF of 0.118 and achieves 19% relative improvement over the state-of-the-art challenge baseline.

Original language	English
Title of host publication	2021 IEEE Spoken Language Technology Workshop (SLT)
Number of pages	7
Publisher	IEEE (Institute of Electrical and Electronics Engineers)
Publication date	25 Mar 2021
Pages	323-329
Article number	9383596
ISBN (Print)	978-1-7281-7067-1
ISBN (Electronic)	978-1-7281-7066-4
DOIs	https://doi.org/10.1109/SLT48900.2021.9383596
Publication status	Published - 25 Mar 2021
Event	2021 IEEE Spoken Language Technology Workshop (SLT) - Shenzhen, China Duration: 19 Jan 2021 → 22 Jan 2021

Conference

Conference	2021 IEEE Spoken Language Technology Workshop (SLT)
Country/Territory	China
City	Shenzhen
Period	19/01/2021 → 22/01/2021

Keywords

Bottleneck feature
Fusion
SdSV challenge 2020
Text-dependent speaker verification
Utterance verification

Access to Document

10.1109/SLT48900.2021.9383596

AUB Link

Search for the material in Aalborg University Library's search engine

Cite this

Sahidullah, M., Sarkar, A. K., Vestman, V., Liu, X., Serizel, R., Kinnunen, T., Tan, Z.-H., & Vincent, E. (2021). UIAI System for Short-Duration Speaker Verification Challenge 2020. In 2021 IEEE Spoken Language Technology Workshop (SLT) (pp. 323-329). Article 9383596 IEEE (Institute of Electrical and Electronics Engineers). https://doi.org/10.1109/SLT48900.2021.9383596

@inproceedings{e70d9078d1014452ad5bc08cf4a0db3e,

title = "UIAI System for Short-Duration Speaker Verification Challenge 2020",

abstract = "In this work, we present the system description of the UIAI entry for the short-duration speaker verification (SdSV) challenge 2020. Our focus is on Task 1 dedicated to text-dependent speaker verification. We investigate different feature extraction and modeling approaches for automatic speaker verification (ASV) and utterance verification (UV). We have also studied different fusion strategies for combining UV and ASV modules. Our primary submission to the challenge is the fusion of seven subsystems which yields a normalized minimum detection cost function (minDCF) of 0.072 and an equal error rate (EER) of 2.14% on the evaluation set. The single system consisting of a pass-phrase identification based model with phone-discriminative bottleneck features gives a normalized minDCF of 0.118 and achieves 19% relative improvement over the state-of-the-art challenge baseline.",

keywords = "Bottleneck feature, Fusion, SdSV challenge 2020, Text-dependent speaker verification, Utterance verification",

author = "Md Sahidullah and Sarkar, {Achintya Kumar} and Ville Vestman and Xuechen Liu and Romain Serizel and Tomi Kinnunen and Zheng-Hua Tan and Emmanuel Vincent",

year = "2021",

month = mar,

day = "25",

doi = "10.1109/SLT48900.2021.9383596",

language = "English",

isbn = "978-1-7281-7067-1",

pages = "323--329",

booktitle = "2021 IEEE Spoken Language Technology Workshop (SLT)",

publisher = "IEEE (Institute of Electrical and Electronics Engineers)",

address = "United States",

note = "2021 IEEE Spoken Language Technology Workshop (SLT) ; Conference date: 19-01-2021 Through 22-01-2021",

}

Sahidullah, M, Sarkar, AK, Vestman, V, Liu, X, Serizel, R, Kinnunen, T, Tan, Z-H & Vincent, E 2021, UIAI System for Short-Duration Speaker Verification Challenge 2020. in 2021 IEEE Spoken Language Technology Workshop (SLT)., 9383596, IEEE (Institute of Electrical and Electronics Engineers), pp. 323-329, 2021 IEEE Spoken Language Technology Workshop (SLT), Shenzhen, China, 19/01/2021. https://doi.org/10.1109/SLT48900.2021.9383596

UIAI System for Short-Duration Speaker Verification Challenge 2020. / Sahidullah, Md; Sarkar, Achintya Kumar; Vestman, Ville et al.
2021 IEEE Spoken Language Technology Workshop (SLT). IEEE (Institute of Electrical and Electronics Engineers), 2021. p. 323-329 9383596.

Research output: Contribution to book/anthology/report/conference proceeding › Article in proceeding › Research › peer-review

TY - GEN

T1 - UIAI System for Short-Duration Speaker Verification Challenge 2020

AU - Sahidullah, Md

AU - Sarkar, Achintya Kumar

AU - Vestman, Ville

AU - Liu, Xuechen

AU - Serizel, Romain

AU - Kinnunen, Tomi

AU - Tan, Zheng-Hua

AU - Vincent, Emmanuel

PY - 2021/3/25

Y1 - 2021/3/25

N2 - In this work, we present the system description of the UIAI entry for the short-duration speaker verification (SdSV) challenge 2020. Our focus is on Task 1 dedicated to text-dependent speaker verification. We investigate different feature extraction and modeling approaches for automatic speaker verification (ASV) and utterance verification (UV). We have also studied different fusion strategies for combining UV and ASV modules. Our primary submission to the challenge is the fusion of seven subsystems which yields a normalized minimum detection cost function (minDCF) of 0.072 and an equal error rate (EER) of 2.14% on the evaluation set. The single system consisting of a pass-phrase identification based model with phone-discriminative bottleneck features gives a normalized minDCF of 0.118 and achieves 19% relative improvement over the state-of-the-art challenge baseline.

AB - In this work, we present the system description of the UIAI entry for the short-duration speaker verification (SdSV) challenge 2020. Our focus is on Task 1 dedicated to text-dependent speaker verification. We investigate different feature extraction and modeling approaches for automatic speaker verification (ASV) and utterance verification (UV). We have also studied different fusion strategies for combining UV and ASV modules. Our primary submission to the challenge is the fusion of seven subsystems which yields a normalized minimum detection cost function (minDCF) of 0.072 and an equal error rate (EER) of 2.14% on the evaluation set. The single system consisting of a pass-phrase identification based model with phone-discriminative bottleneck features gives a normalized minDCF of 0.118 and achieves 19% relative improvement over the state-of-the-art challenge baseline.

KW - Bottleneck feature

KW - Fusion

KW - SdSV challenge 2020

KW - Text-dependent speaker verification

KW - Utterance verification

UR - http://www.scopus.com/inward/record.url?scp=85101904954&partnerID=8YFLogxK

U2 - 10.1109/SLT48900.2021.9383596

DO - 10.1109/SLT48900.2021.9383596

M3 - Article in proceeding

SN - 978-1-7281-7067-1

SP - 323

EP - 329

BT - 2021 IEEE Spoken Language Technology Workshop (SLT)

PB - IEEE (Institute of Electrical and Electronics Engineers)

T2 - 2021 IEEE Spoken Language Technology Workshop (SLT)

Y2 - 19 January 2021 through 22 January 2021

ER -

UIAI System for Short-Duration Speaker Verification Challenge 2020

Abstract

Conference

Keywords

Access to Document

AUB Link

Other files and links

Fingerprint

Cite this