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 proceedingArticle in proceedingResearchpeer-review

2 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 languageEnglish
Title of host publication2021 IEEE Spoken Language Technology Workshop (SLT)
Number of pages7
PublisherIEEE
Publication date25 Mar 2021
Pages323-329
Article number9383596
ISBN (Print)978-1-7281-7067-1
ISBN (Electronic)978-1-7281-7066-4
DOIs
Publication statusPublished - 25 Mar 2021
Event2021 IEEE Spoken Language Technology Workshop (SLT) - Shenzhen, China
Duration: 19 Jan 202122 Jan 2021

Conference

Conference2021 IEEE Spoken Language Technology Workshop (SLT)
CountryChina
CityShenzhen
Period19/01/202122/01/2021

Keywords

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

Fingerprint

Dive into the research topics of 'UIAI System for Short-Duration Speaker Verification Challenge 2020'. Together they form a unique fingerprint.

Cite this