Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric Bias
This paper describes a new unsupervised machine-learning method for simultaneous phoneme and word discovery from multiple speakers. Phoneme and word discovery from multiple speakers is a more challenging problem than that from one speaker, because the speech signals from different speakers exhibit d...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Frontiers Media S.A.
2019-10-01
|
Series: | Frontiers in Robotics and AI |
Subjects: | |
Online Access: | https://www.frontiersin.org/article/10.3389/frobt.2019.00092/full |
id |
doaj-4b932ba456b74b6bb492cd16ee1b14af |
---|---|
record_format |
Article |
spelling |
doaj-4b932ba456b74b6bb492cd16ee1b14af2020-11-25T01:34:57ZengFrontiers Media S.A.Frontiers in Robotics and AI2296-91442019-10-01610.3389/frobt.2019.00092458555Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric BiasRyo NakashimaRyo OzakiTadahiro TaniguchiThis paper describes a new unsupervised machine-learning method for simultaneous phoneme and word discovery from multiple speakers. Phoneme and word discovery from multiple speakers is a more challenging problem than that from one speaker, because the speech signals from different speakers exhibit different acoustic features. The existing method, a nonparametric Bayesian double articulation analyzer (NPB-DAA) with deep sparse autoencoder (DSAE) only performed phoneme and word discovery from a single speaker. Extending NPB-DAA with DSAE to a multi-speaker scenario is, therefore, the research problem of this paper.This paper proposes the employment of a DSAE with parametric bias in the hidden layer (DSAE-PBHL) as a feature extractor for unsupervised phoneme and word discovery. DSAE-PBHL is designed to subtract speaker-dependent acoustic features and speaker-independent features by introducing parametric bias input to the DSAE hidden layer. An experiment demonstrated that DSAE-PBHL could subtract distributed representations of acoustic signals, enabling extraction based on the types of phonemes rather than the speakers. Another experiment demonstrated that a combination of NPB-DAA and DSAE-PBHL outperformed other available methods accomplishing phoneme and word discovery tasks involving speech signals with Japanese vowel sequences from multiple speakers.https://www.frontiersin.org/article/10.3389/frobt.2019.00092/fullword discoveryphoneme discoveryparametric biasBayesian modelneural network |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Ryo Nakashima Ryo Ozaki Tadahiro Taniguchi |
spellingShingle |
Ryo Nakashima Ryo Ozaki Tadahiro Taniguchi Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric Bias Frontiers in Robotics and AI word discovery phoneme discovery parametric bias Bayesian model neural network |
author_facet |
Ryo Nakashima Ryo Ozaki Tadahiro Taniguchi |
author_sort |
Ryo Nakashima |
title |
Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric Bias |
title_short |
Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric Bias |
title_full |
Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric Bias |
title_fullStr |
Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric Bias |
title_full_unstemmed |
Unsupervised Phoneme and Word Discovery From Multiple Speakers Using Double Articulation Analyzer and Neural Network With Parametric Bias |
title_sort |
unsupervised phoneme and word discovery from multiple speakers using double articulation analyzer and neural network with parametric bias |
publisher |
Frontiers Media S.A. |
series |
Frontiers in Robotics and AI |
issn |
2296-9144 |
publishDate |
2019-10-01 |
description |
This paper describes a new unsupervised machine-learning method for simultaneous phoneme and word discovery from multiple speakers. Phoneme and word discovery from multiple speakers is a more challenging problem than that from one speaker, because the speech signals from different speakers exhibit different acoustic features. The existing method, a nonparametric Bayesian double articulation analyzer (NPB-DAA) with deep sparse autoencoder (DSAE) only performed phoneme and word discovery from a single speaker. Extending NPB-DAA with DSAE to a multi-speaker scenario is, therefore, the research problem of this paper.This paper proposes the employment of a DSAE with parametric bias in the hidden layer (DSAE-PBHL) as a feature extractor for unsupervised phoneme and word discovery. DSAE-PBHL is designed to subtract speaker-dependent acoustic features and speaker-independent features by introducing parametric bias input to the DSAE hidden layer. An experiment demonstrated that DSAE-PBHL could subtract distributed representations of acoustic signals, enabling extraction based on the types of phonemes rather than the speakers. Another experiment demonstrated that a combination of NPB-DAA and DSAE-PBHL outperformed other available methods accomplishing phoneme and word discovery tasks involving speech signals with Japanese vowel sequences from multiple speakers. |
topic |
word discovery phoneme discovery parametric bias Bayesian model neural network |
url |
https://www.frontiersin.org/article/10.3389/frobt.2019.00092/full |
work_keys_str_mv |
AT ryonakashima unsupervisedphonemeandworddiscoveryfrommultiplespeakersusingdoublearticulationanalyzerandneuralnetworkwithparametricbias AT ryoozaki unsupervisedphonemeandworddiscoveryfrommultiplespeakersusingdoublearticulationanalyzerandneuralnetworkwithparametricbias AT tadahirotaniguchi unsupervisedphonemeandworddiscoveryfrommultiplespeakersusingdoublearticulationanalyzerandneuralnetworkwithparametricbias |
_version_ |
1725069390305558528 |