Publication result detail

Analyzing speaker verification embedding extractors and back-ends under language and channel mismatch

SILNOVA, A.; STAFYLAKIS, T.; MOŠNER, L.; PLCHOT, O.; ROHDIN, J.; MATĚJKA, P.; BURGET, L.; GLEMBEK, O.; BRUMMER, J.

Original Title

Analyzing speaker verification embedding extractors and back-ends under language and channel mismatch

English Title

Analyzing speaker verification embedding extractors and back-ends under language and channel mismatch

Type

Paper in proceedings outside WoS and Scopus

Original Abstract

In this paper, we analyze the behavior and performance ofspeaker embeddings and the back-end scoring model under domainand language mismatch. We present our findings regardingResNet-based speaker embedding architectures and showthat reduced temporal stride yields improved performance. Wethen consider a PLDA back-end and show how a combinationof small speaker subspace, language-dependent PLDA mixture,and nuisance-attribute projection can have a drastic impact onthe performance of the system. Besides, we present an efficientway of scoring and fusing class posterior logit vectors recentlyshown to perform well on speaker verification task. The experimentsare performed using the NIST SRE 2021 setup.

English abstract

In this paper, we analyze the behavior and performance ofspeaker embeddings and the back-end scoring model under domainand language mismatch. We present our findings regardingResNet-based speaker embedding architectures and showthat reduced temporal stride yields improved performance. Wethen consider a PLDA back-end and show how a combinationof small speaker subspace, language-dependent PLDA mixture,and nuisance-attribute projection can have a drastic impact onthe performance of the system. Besides, we present an efficientway of scoring and fusing class posterior logit vectors recentlyshown to perform well on speaker verification task. The experimentsare performed using the NIST SRE 2021 setup.

Keywords

speaker, verification, embedding

Key words in English

speaker, verification, embedding

Authors

SILNOVA, A.; STAFYLAKIS, T.; MOŠNER, L.; PLCHOT, O.; ROHDIN, J.; MATĚJKA, P.; BURGET, L.; GLEMBEK, O.; BRUMMER, J.

Released

28.06.2022

Publisher

International Speech Communication Association

Location

Beijing

Book

Proceedings of The Speaker and Language Recognition Workshop (Odyssey 2022)

Pages from

9

Pages to

16

Pages count

8

URL

https://www.isca-speech.org/archive/pdfs/odyssey_2022/silnova22_odyssey.pdf

BibTex

@inproceedings{BUT179660,
  author="Anna {Silnova} and Themos {Stafylakis} and Ladislav {Mošner} and Oldřich {Plchot} and Johan Andréas {Rohdin} and Pavel {Matějka} and Lukáš {Burget} and Ondřej {Glembek} and Johan Nikolaas Langenhoven {Brummer}",
  title="Analyzing speaker verification embedding extractors and back-ends under language and channel mismatch",
  booktitle="Proceedings of The Speaker and Language Recognition Workshop (Odyssey 2022)",
  year="2022",
  pages="9--16",
  publisher="International Speech Communication Association",
  address="Beijing",
  doi="10.21437/Odyssey.2022-2",
  url="https://www.isca-speech.org/archive/pdfs/odyssey_2022/silnova22_odyssey.pdf"
}

Documents

silnova_odyssey2022_analysis