Publication Details
BUT system for DIHARD Speech Diarization Challenge 2018
Landini Federico Nicolás, Ph.D. (RG SPEECH)
Burget Lukáš, doc. Ing., Ph.D. (DCGM)
Rohdin Johan Andréas, M.Sc., Ph.D. (DCGM)
Silnova Anna, M.Sc., Ph.D. (DCGM)
Žmolíková Kateřina, Ing., Ph.D. (FIT)
Novotný Ondřej, Ing., Ph.D.
Veselý Karel, Ing., Ph.D. (DCGM)
Glembek Ondřej, Ing., Ph.D.
Plchot Oldřich, Ing., Ph.D. (DCGM)
Mošner Ladislav, Ing. (DCGM)
Matějka Pavel, Ing., Ph.D. (DCGM)
Speaker Diarization, Variational Bayes, HMM,i-vector, x-vector, Overlapped speech, DIHARD
This paper presents the approach developed by the BUT teamfor the first DIHARD speech diarization challenge, which isbased on our Bayesian Hidden Markov Model with eigenvoicepriors system. Besides the description of the approach, we providea brief analysis of different techniques and data processingmethods tested on the development set. We also introducea simple attempt for overlapped speech detection that we usedfor attaining cleaner speaker models and reassigning overlappedspeech to multiple speakers. Finally, we present results obtainedon the evaluation set and discuss findings we made during thedevelopment phase and with the help of the DIHARD leaderboardfeedback.
@inproceedings{BUT155100,
author="Mireia {Diez Sánchez} and Federico Nicolás {Landini} and Lukáš {Burget} and Johan Andréas {Rohdin} and Anna {Silnova} and Kateřina {Žmolíková} and Ondřej {Novotný} and Karel {Veselý} and Ondřej {Glembek} and Oldřich {Plchot} and Ladislav {Mošner} and Pavel {Matějka}",
title="BUT system for DIHARD Speech Diarization Challenge 2018",
booktitle="Proceedings of Interspeech 2018",
year="2018",
journal="Proceedings of Interspeech",
volume="2018",
number="9",
pages="2798--2802",
publisher="International Speech Communication Association",
address="Hyderabad",
doi="10.21437/Interspeech.2018-1749",
issn="1990-9772",
url="https://www.isca-speech.org/archive/Interspeech_2018/abstracts/1749.html"
}