Publication Details
But System for the Second Dihard Speech Diarization Challenge
Wang Shuai
Diez Sánchez Mireia, M.Sc., Ph.D. (DCGM)
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Matějka Pavel, Ing., Ph.D.
Žmolíková Kateřina, Ing., Ph.D. (FIT)
Mošner Ladislav, Ing. (DCGM)
Silnova Anna, M.Sc., Ph.D. (DCGM)
Plchot Oldřich, Ing., Ph.D. (DCGM)
Novotný Ondřej, Ing., Ph.D.
Zeinali Hossein, Ph.D. (DCGM)
Rohdin Johan Andréas, M.Sc., Ph.D. (DCGM)
Speaker Diarization, Variational Bayes, HMM, DIHARD, CHiME
This paper describes the winning systems developed by theBUT team for the four tracks of the Second DIHARD SpeechDiarization Challenge. For tracks 1 and 2 the systems weremainly based on performing agglomerative hierarchical clustering(AHC) of x-vectors, followed by another x-vectorclustering based on Bayes hidden Markov model and variationalBayes inference. We provide a comparison of theimprovement given by each step and share the implementationof the core of the system. For tracks 3 and 4 withrecordings from the Fifth CHiME Challenge, we exploreddifferent approaches for doing multi-channel diarization andour best performance was obtained when applying AHC onthe fusion of per channel probabilistic linear discriminantanalysis scores.
@inproceedings{BUT163962,
author="Federico Nicolás {Landini} and Shuai {Wang} and Mireia {Diez Sánchez} and Lukáš {Burget} and Pavel {Matějka} and Kateřina {Žmolíková} and Ladislav {Mošner} and Anna {Silnova} and Oldřich {Plchot} and Ondřej {Novotný} and Hossein {Zeinali} and Johan Andréas {Rohdin}",
title="But System for the Second Dihard Speech Diarization Challenge",
booktitle="ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",
year="2020",
pages="6529--6533",
publisher="IEEE Signal Processing Society",
address="Barcelona",
doi="10.1109/ICASSP40776.2020.9054251",
isbn="978-1-5090-6631-5",
url="https://ieeexplore.ieee.org/document/9054251"
}