Detail výsledku

An attention-based backend allowing efficient fine-tuning of transformer models for speaker verification

PENG, J.; PLCHOT, O.; STAFYLAKIS, T.; MOŠNER, L.; BURGET, L.; ČERNOCKÝ, J. An attention-based backend allowing efficient fine-tuning of transformer models for speaker verification. In 2022 IEEE Spoken Language Technology Workshop, SLT 2022 - Proceedings. Doha: IEEE Signal Processing Society, 2023. p. 555-562. ISBN: 978-1-6654-7189-3.
Typ
článek ve sborníku konference
Jazyk
anglicky
Autoři
Peng Junyi, UPGM (FIT)
Plchot Oldřich, Ing., Ph.D., UPGM (FIT)
Stafylakis Themos
Mošner Ladislav, Ing., UPGM (FIT)
Burget Lukáš, doc. Ing., Ph.D., UPGM (FIT)
Černocký Jan, prof. Dr. Ing., UPGM (FIT)
Abstrakt

In recent years, self-supervised learning paradigm has received extensive
attention due to its great success in various down-stream tasks.
However, the fine-tuning strategies for adapting those pre-trained
models to speaker verification task have yet to be fully explored. In
this paper, we analyze several feature extraction approaches built on
top of a pre-trained model, as well as regularization and a learning
rate scheduler to stabilize the fine-tuning process and further boost
performance: multi-head factorized attentive pooling is proposed
to factorize the comparison of speaker representations into multiple
phonetic clusters. We regularize towards the parameters of the pretrained
model and we set different learning rates for each layer of the
pre-trained model during fine-tuning. The experimental results show
our method can significantly shorten the training time to 4 hours
and achieve SOTA performance: 0.59%, 0.79% and 1.77% EER on
Vox1-O, Vox1-E and Vox1-H, respectively.

Klíčová slova

Pre-trained model, fine-tuning strategy, speaker verification, attentive pooling

URL
Rok
2023
Strany
555–562
Sborník
2022 IEEE Spoken Language Technology Workshop, SLT 2022 - Proceedings
Konference
IEEE SPOKEN LANGUAGE TECHNOLOGY WORKSHOP, SLT
ISBN
978-1-6654-7189-3
Vydavatel
IEEE Signal Processing Society
Místo
Doha
DOI
10.1109/SLT54892.2023.10022775
UT WoS
000968851900075
EID Scopus
2-s2.0-85147794754
BibTeX 
@inproceedings{BUT185120,
  author="Junyi {Peng} and Oldřich {Plchot} and Themos {Stafylakis} and Ladislav {Mošner} and Lukáš {Burget} and Jan {Černocký}",
  title="An attention-based backend allowing efficient fine-tuning of transformer models for speaker verification",
  booktitle="2022 IEEE Spoken Language Technology Workshop, SLT 2022 - Proceedings",
  year="2023",
  pages="555--562",
  publisher="IEEE Signal Processing Society",
  address="Doha",
  doi="10.1109/SLT54892.2023.10022775",
  isbn="978-1-6654-7189-3",
  url="https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10022775"
}
Soubory
Projekty
Multi-lingualita v řečových technologiích, MŠMT, INTER-EXCELLENCE - Podprogram INTER-ACTION, LTAIN19087, zahájení: 2020-01-01, ukončení: 2023-08-31, ukončen
Neuronové reprezentace v multimodálním a mnohojazyčném modelování, GAČR, Grantové projekty exelence v základním výzkumu EXPRO - 2019, GX19-26934X, zahájení: 2019-01-01, ukončení: 2023-12-31, ukončen
Výměny pro výzkum řeči a technologií, EU, Horizon 2020, zahájení: 2021-01-01, ukončení: 2025-12-31, řešení
Výzkumné skupiny
Výzkumná skupina dolování dat z řeči BUT Speech@FIT (VZ SPEECH)
Pracoviště
Ústav počítačové grafiky a multimédií (UPGM)
Nahoru