View in EDS

Uncertainty-based learning of acoustic models from noisy data

Saved in:
Bibliographic Details
Title: Uncertainty-based learning of acoustic models from noisy data
Authors: Ozerov, Alexey1 alexey.ozerov@technicolor.com, Lagrange, Mathieu2 mathieu.lagrange@ircam.fr, Vincent, Emmanuel3 emmanuel.vincent@inria.fr
Source: Computer Speech & Language. May2013, Vol. 27 Issue 3, p874-894. 21p.
Subjects: Acoustic models, Computational learning theory, Gaussian distribution, Decoding algorithms, Mathematical optimization, Gaussian mixture models, Hidden Markov models, Regression analysis
Abstract: Abstract: We consider the problem of acoustic modeling of noisy speech data, where the uncertainty over the data is given by a Gaussian distribution. While this uncertainty has been exploited at the decoding stage via uncertainty decoding, its usage at the training stage remains limited to static model adaptation. We introduce a new expectation maximization (EM) based technique, which we call uncertainty training, that allows us to train Gaussian mixture models (GMMs) or hidden Markov models (HMMs) directly from noisy data with dynamic uncertainty. We evaluate the potential of this technique for a GMM-based speaker recognition task on speech data corrupted by real-world domestic background noise, using a state-of-the-art signal enhancement technique and various uncertainty estimation techniques as a front-end. Compared to conventional training, the proposed training algorithm results in 3–4% absolute improvement in speaker recognition accuracy by training from either matched, unmatched or multi-condition noisy data. This algorithm is also applicable with minor modifications to maximum a posteriori (MAP) or maximum likelihood linear regression (MLLR) acoustic model adaptation from noisy data and to other data than audio. [Copyright &y& Elsevier]
Copyright of Computer Speech & Language is the property of Academic Press Inc. and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
FullText Text:
  Availability: 0
Header DbId: egs
DbLabel: Engineering Source
An: 85264541
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Uncertainty-based learning of acoustic models from noisy data
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Ozerov%2C+Alexey%22">Ozerov, Alexey</searchLink><relatesTo>1</relatesTo><i> alexey.ozerov@technicolor.com</i><br /><searchLink fieldCode="AR" term="%22Lagrange%2C+Mathieu%22">Lagrange, Mathieu</searchLink><relatesTo>2</relatesTo><i> mathieu.lagrange@ircam.fr</i><br /><searchLink fieldCode="AR" term="%22Vincent%2C+Emmanuel%22">Vincent, Emmanuel</searchLink><relatesTo>3</relatesTo><i> emmanuel.vincent@inria.fr</i>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Computer+Speech+%26+Language%22">Computer Speech & Language</searchLink>. May2013, Vol. 27 Issue 3, p874-894. 21p.
– Name: Subject
  Label: Subjects
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Acoustic+models%22">Acoustic models</searchLink><br /><searchLink fieldCode="DE" term="%22Computational+learning+theory%22">Computational learning theory</searchLink><br /><searchLink fieldCode="DE" term="%22Gaussian+distribution%22">Gaussian distribution</searchLink><br /><searchLink fieldCode="DE" term="%22Decoding+algorithms%22">Decoding algorithms</searchLink><br /><searchLink fieldCode="DE" term="%22Mathematical+optimization%22">Mathematical optimization</searchLink><br /><searchLink fieldCode="DE" term="%22Gaussian+mixture+models%22">Gaussian mixture models</searchLink><br /><searchLink fieldCode="DE" term="%22Hidden+Markov+models%22">Hidden Markov models</searchLink><br /><searchLink fieldCode="DE" term="%22Regression+analysis%22">Regression analysis</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Abstract: We consider the problem of acoustic modeling of noisy speech data, where the uncertainty over the data is given by a Gaussian distribution. While this uncertainty has been exploited at the decoding stage via uncertainty decoding, its usage at the training stage remains limited to static model adaptation. We introduce a new expectation maximization (EM) based technique, which we call uncertainty training, that allows us to train Gaussian mixture models (GMMs) or hidden Markov models (HMMs) directly from noisy data with dynamic uncertainty. We evaluate the potential of this technique for a GMM-based speaker recognition task on speech data corrupted by real-world domestic background noise, using a state-of-the-art signal enhancement technique and various uncertainty estimation techniques as a front-end. Compared to conventional training, the proposed training algorithm results in 3–4% absolute improvement in speaker recognition accuracy by training from either matched, unmatched or multi-condition noisy data. This algorithm is also applicable with minor modifications to maximum a posteriori (MAP) or maximum likelihood linear regression (MLLR) acoustic model adaptation from noisy data and to other data than audio. [Copyright &y& Elsevier]
– Name: AbstractSuppliedCopyright
  Label:
  Group: Ab
  Data: <i>Copyright of Computer Speech & Language is the property of Academic Press Inc. and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.)
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=85264541
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1016/j.csl.2012.07.002
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 21
        StartPage: 874
    Subjects:
      – SubjectFull: Acoustic models
        Type: general
      – SubjectFull: Computational learning theory
        Type: general
      – SubjectFull: Gaussian distribution
        Type: general
      – SubjectFull: Decoding algorithms
        Type: general
      – SubjectFull: Mathematical optimization
        Type: general
      – SubjectFull: Gaussian mixture models
        Type: general
      – SubjectFull: Hidden Markov models
        Type: general
      – SubjectFull: Regression analysis
        Type: general
    Titles:
      – TitleFull: Uncertainty-based learning of acoustic models from noisy data
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Ozerov, Alexey
      – PersonEntity:
          Name:
            NameFull: Lagrange, Mathieu
      – PersonEntity:
          Name:
            NameFull: Vincent, Emmanuel
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 05
              Text: May2013
              Type: published
              Y: 2013
          Identifiers:
            – Type: issn-print
              Value: 08852308
          Numbering:
            – Type: volume
              Value: 27
            – Type: issue
              Value: 3
          Titles:
            – TitleFull: Computer Speech & Language
              Type: main
ResultId 1