Detail publikačního výsledku

Optimizing Bayesian Hmm Based X-Vector Clustering for the Second Dihard Speech Diarization Challenge

DIEZ SÁNCHEZ, M.; BURGET, L.; LANDINI, F.; WANG, S.; ČERNOCKÝ, J.

Originální název

Optimizing Bayesian Hmm Based X-Vector Clustering for the Second Dihard Speech Diarization Challenge

Anglický název

Optimizing Bayesian Hmm Based X-Vector Clustering for the Second Dihard Speech Diarization Challenge

Druh

Stať ve sborníku v databázi WoS či Scopus

Originální abstrakt

This paper presents an analysis of our diarization systemwinning the second DIHARD speech diarization challenge,track 1. This system is based on clustering x-vector speakerembeddings extracted every 0.25s from short segments of theinput recording. In this paper, we focus on the two x-vectorclustering methods employed, namely Agglomerative HierarchicalClustering followed by a clustering based on BayesianHidden Markov Model (BHMM). Even though the systemsubmitted to the challenge had further post-processing steps,we will show that using this BHMM solely is enough toachieve the best performance in the challenge. The analysiswill show improvements achieved by optimizing individualprocessing steps, including a simple procedure to effectivelyperform "domain adaptation" by Probabilistic LinearDiscriminant Analysis model interpolation. All experimentsare performed in the DIHARD II evaluation framework.

Anglický abstrakt

This paper presents an analysis of our diarization systemwinning the second DIHARD speech diarization challenge,track 1. This system is based on clustering x-vector speakerembeddings extracted every 0.25s from short segments of theinput recording. In this paper, we focus on the two x-vectorclustering methods employed, namely Agglomerative HierarchicalClustering followed by a clustering based on BayesianHidden Markov Model (BHMM). Even though the systemsubmitted to the challenge had further post-processing steps,we will show that using this BHMM solely is enough toachieve the best performance in the challenge. The analysiswill show improvements achieved by optimizing individualprocessing steps, including a simple procedure to effectivelyperform "domain adaptation" by Probabilistic LinearDiscriminant Analysis model interpolation. All experimentsare performed in the DIHARD II evaluation framework.

Klíčová slova

Speaker Diarization, Variational Bayes, HMM, x-vector, DIHARD

Klíčová slova v angličtině

Speaker Diarization, Variational Bayes, HMM, x-vector, DIHARD

Autoři

DIEZ SÁNCHEZ, M.; BURGET, L.; LANDINI, F.; WANG, S.; ČERNOCKÝ, J.

Rok RIV

2021

Vydáno

04.05.2020

Nakladatel

IEEE Signal Processing Society

Místo

Barcelona

ISBN

978-1-5090-6631-5

Kniha

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

Strany od

6519

Strany do

6523

Strany počet

5

URL

https://ieeexplore.ieee.org/document/9053982

BibTex

@inproceedings{BUT163963,
  author="Mireia {Diez Sánchez} and Lukáš {Burget} and Federico Nicolás {Landini} and Shuai {Wang} and Jan {Černocký}",
  title="Optimizing Bayesian Hmm Based X-Vector Clustering for the Second Dihard Speech Diarization Challenge",
  booktitle="ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",
  year="2020",
  pages="6519--6523",
  publisher="IEEE Signal Processing Society",
  address="Barcelona",
  doi="10.1109/ICASSP40776.2020.9053982",
  isbn="978-1-5090-6631-5",
  url="https://ieeexplore.ieee.org/document/9053982"
}

Dokumenty

diez_icassp2020_09053982