Detail publikace

Speaker Diarization based on Bayesian HMM with Eigenvoice Priors

DIEZ SÁNCHEZ, M. BURGET, L. MATĚJKA, P.

Originální název

Speaker Diarization based on Bayesian HMM with Eigenvoice Priors

Anglický název

Speaker Diarization based on Bayesian HMM with Eigenvoice Priors

Jazyk

en

Originální abstrakt

Nowadays, most speaker diarization methods address the task in two steps: segmentation of the input conversation into (preferably) speaker homogeneous segments, and clustering. Generally, different models and techniques are used for the two steps. In this paper we present a very elegant approach where a straightforward and efficient Variational Bayes (VB) inference in a single probabilistic model addresses the complete SD problem. Our model is a Bayesian Hidden Markov Model, in which states represent speaker specific distributions and transitions between states represent speaker turns. As in the ivector or JFA models, speaker distributions are modeled by GMMs with parameters constrained by eigenvoice priors. This allows to robustly estimate the speaker models from very short speech segments. The model, which was released as open source code and has already been used by several labs, is fully described for the first time in this paper. We present results and the system is compared and combined with other state-of-the-art approaches. The model provides the best results reported so far on the CALLHOME dataset.

Anglický abstrakt

Nowadays, most speaker diarization methods address the task in two steps: segmentation of the input conversation into (preferably) speaker homogeneous segments, and clustering. Generally, different models and techniques are used for the two steps. In this paper we present a very elegant approach where a straightforward and efficient Variational Bayes (VB) inference in a single probabilistic model addresses the complete SD problem. Our model is a Bayesian Hidden Markov Model, in which states represent speaker specific distributions and transitions between states represent speaker turns. As in the ivector or JFA models, speaker distributions are modeled by GMMs with parameters constrained by eigenvoice priors. This allows to robustly estimate the speaker models from very short speech segments. The model, which was released as open source code and has already been used by several labs, is fully described for the first time in this paper. We present results and the system is compared and combined with other state-of-the-art approaches. The model provides the best results reported so far on the CALLHOME dataset.

Dokumenty

BibTex


@inproceedings{BUT155067,
  author="Mireia {Diez Sánchez} and Lukáš {Burget} and Pavel {Matějka}",
  title="Speaker Diarization based on Bayesian HMM with Eigenvoice Priors",
  annote="Nowadays, most speaker diarization methods address the task in two steps:
segmentation of the input conversation into (preferably) speaker homogeneous
segments, and clustering. Generally, different models and techniques are used for
the two steps. In this paper we present a very elegant approach where a
straightforward and efficient Variational Bayes (VB) inference in a single
probabilistic model addresses the complete SD problem. Our model is a Bayesian
Hidden Markov Model, in which states represent speaker specific distributions and
transitions between states represent speaker turns. As in the ivector or JFA
models, speaker distributions are modeled by GMMs with parameters constrained by
eigenvoice priors. This allows to robustly estimate the speaker models from very
short speech segments. The model, which was released as open source code and has
already been used by several labs, is fully described for the first time in this
paper. We present results and the system is compared and combined with other
state-of-the-art approaches. The model provides the best results reported so far
on the CALLHOME dataset.",
  address="International Speech Communication Association",
  booktitle="Proceedings of Odyssey 2018",
  chapter="155067",
  doi="10.21437/Odyssey.2018-21",
  edition="NEUVEDEN",
  howpublished="online",
  institution="International Speech Communication Association",
  number="6",
  year="2018",
  month="june",
  pages="147--154",
  publisher="International Speech Communication Association",
  type="conference paper"
}