Detail publikace

Discriminatively Re-trained i-Vector Extractor For Speaker Recognition

NOVOTNÝ, O. PLCHOT, O. GLEMBEK, O. BURGET, L. MATĚJKA, P.

Originální název

Discriminatively Re-trained i-Vector Extractor For Speaker Recognition

Anglický název

Discriminatively Re-trained i-Vector Extractor For Speaker Recognition

Jazyk

en

Originální abstrakt

In this work we revisit discriminative training of the i-vector extractor component in the standard speaker verification (SV) system. The motivation of our research lies in the robustness and stability of this large generative model, which we want to preserve, and focus its power towards any intended SV task. We show that after generative initialization of the i-vector extractor, we can further refine it with discriminative training and obtain i-vectors that lead to better performance on various benchmarks representing different acoustic domains.

Anglický abstrakt

In this work we revisit discriminative training of the i-vector extractor component in the standard speaker verification (SV) system. The motivation of our research lies in the robustness and stability of this large generative model, which we want to preserve, and focus its power towards any intended SV task. We show that after generative initialization of the i-vector extractor, we can further refine it with discriminative training and obtain i-vectors that lead to better performance on various benchmarks representing different acoustic domains.

Dokumenty

BibTex


@inproceedings{BUT160000,
  author="Ondřej {Novotný} and Oldřich {Plchot} and Ondřej {Glembek} and Lukáš {Burget} and Pavel {Matějka}",
  title="Discriminatively Re-trained i-Vector Extractor For Speaker Recognition",
  annote="In this work we revisit discriminative training of the i-vector extractor
component in the standard speaker verification (SV) system. The motivation of our
research lies in the robustness and stability of this large generative model,
which we want to preserve, and focus its power towards any intended SV task. We
show that after generative initialization of the i-vector extractor, we can
further refine it with discriminative training and obtain i-vectors that lead to
better performance on various benchmarks representing different acoustic
domains.",
  address="IEEE Signal Processing Society",
  booktitle="Proceedings of 2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)",
  chapter="160000",
  doi="10.1109/ICASSP.2019.8682590",
  edition="NEUVEDEN",
  howpublished="electronic, physical medium",
  institution="IEEE Signal Processing Society",
  year="2019",
  month="may",
  pages="6031--6035",
  publisher="IEEE Signal Processing Society",
  type="conference paper"
}