Result Details
BUT systems and analyses for the ASVspoof 5 Challenge
ZHANG, L.
Plchot Oldřich, Ing., Ph.D., DCGM (FIT)
Staněk Vojtěch, Ing., DCGM (FIT)
Mihola David, Ing., DCGM (FIT)
Peng Junyi, DCGM (FIT)
Stafylakis Themos
BEVERAKI, D.
Silnova Anna, M.Sc., Ph.D., DCGM (FIT)
Brukner Jan, Ing., DCGM (FIT)
Burget Lukáš, doc. Ing., Ph.D., DCGM (FIT)
This paper describes the BUT submitted systems for the ASVspoof5 challenge, along with analyses. For the conventional deepfake detection task, we use ResNet18 and self-supervised models for the closed and open conditions, respectively. In addition, we analyze and visualize different combinations of speaker information and spoofing information as label schemes for training. For spoofing-robust automatic speaker verification (SASV), we introduce effective priors and propose using logistic regression to jointly train affine transformations of the countermeasure scores and the automatic speaker verification scores in such a way that the SASV LLR is optimized.
spoof, challenge, speaker verification
@inproceedings{BUT194004,
author="ROHDIN, J. and ZHANG, L. and PLCHOT, O. and STANĚK, V. and MIHOLA, D. and PENG, J. and STAFYLAKIS, T. and BEVERAKI, D. and SILNOVA, A. and BRUKNER, J. and BURGET, L.",
title="BUT systems and analyses for the ASVspoof 5 Challenge",
booktitle="Proceedings of ASV spoof 2024 Workshop",
year="2024",
pages="24--31",
publisher="International Speech Communication Association",
address="Kos Island",
doi="10.21437/ASVspoof.2024-4",
url="https://www.isca-archive.org/asvspoof_2024/rohdin24_asvspoof.pdf"
}
Tools To Combat Voice DeepFakes, MV, Programu bezpečnostního výzkumu ČR 2021-2026: vývoj, testování a evaluace nových bezpečnostních technologií (SECTECH) - II. veřejná soutěž, VB02000060, start: 2024-01-01, end: 2026-12-31, running
Speech Data Mining Research Group BUT Speech@FIT (RG SPEECH)