Publication Details

Probabilistic embeddings for speaker diarization

SILNOVA, A.; BRUMMER, J.; ROHDIN, J.; STAFYLAKIS, T.; BURGET, L. Probabilistic embeddings for speaker diarization. Proceedings of Odyssey 2020 The Speaker and Language Recognition Workshop. Proceedings of Odyssey: The Speaker and Language Recognition Workshop Odyssey 2014, Joensuu, Finland. Tokyo: International Speech Communication Association, 2020. p. 24-31. ISSN: 2312-2846.
Czech title
Pravděpodobnostní embeddingy pro diarizaci řečníků
Type
conference paper
Language
English
Authors
Silnova Anna, M.Sc., Ph.D. (DCGM)
Brummer Johan Nikolaas Langenhoven, Dr.
Rohdin Johan Andréas, M.Sc., Ph.D. (DCGM)
Stafylakis Themos
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
URL
Keywords

probabilistic embeddings, speaker diarization

Abstract

Speaker embeddings (x-vectors) extracted from very short segments of speech have
recently been shown to give competitive performance in speaker diarization. We
generalize this recipe by extracting from each speech segment, in parallel with
the x-vector, also a diagonal precision matrix, thus providing a path for the
propagation of information about the quality of the speech segment into a PLDA
scoring backend. These precisions quantify the uncertainty about what the values
of the embeddings might have been if they had been extracted from high quality
speech segments. The proposed probabilistic embeddings (x-vectors with
precisions) are interfaced with the PLDA model by treating the x-vectors as
hidden variables and marginalizing them out. We apply the proposed probabilistic
embeddings as input to an agglomerative hierarchical clustering (AHC) algorithm
to do diarization in the DIHARD19 evaluation set. We compute the full PLDA
likelihood by the book for each clustering hypothesis that is considered by AHC.
We do joint discriminative training of the PLDA parameters and of the
probabilistic x-vector extractor. We demonstrate accuracy gains relative to
a baseline AHC algorithm, applied to traditional xvectors (without uncertainty),
and which uses averaging of binary log-likelihood-ratios, rather than by-the-book
scoring.

Published
2020
Pages
24–31
Journal
Proceedings of Odyssey: The Speaker and Language Recognition Workshop Odyssey 2014, Joensuu, Finland, vol. 2020, no. 11, ISSN 2312-2846
Proceedings
Proceedings of Odyssey 2020 The Speaker and Language Recognition Workshop
Conference
Odyssey 2020: The Speaker and Language Recognition Workshop, Tokyo, JP
Publisher
International Speech Communication Association
Place
Tokyo
DOI
BibTeX
@inproceedings{BUT164068,
  author="Anna {Silnova} and Johan Nikolaas Langenhoven {Brummer} and Johan Andréas {Rohdin} and Themos {Stafylakis} and Lukáš {Burget}",
  title="Probabilistic embeddings for speaker diarization",
  booktitle="Proceedings of Odyssey 2020 The Speaker and Language Recognition Workshop",
  year="2020",
  journal="Proceedings of Odyssey: The Speaker and Language Recognition Workshop Odyssey 2014, Joensuu, Finland",
  volume="2020",
  number="11",
  pages="24--31",
  publisher="International Speech Communication Association",
  address="Tokyo",
  doi="10.21437/Odyssey.2020-4",
  issn="2312-2846",
  url="https://www.isca-speech.org/archive/Odyssey_2020/abstracts/75.html"
}
Files
Back to top