A General Joint Additive and Convolutive Bias Compensation Approach Applied to Noisy Lombard Speech Recognition
Identifieur interne : 00B649 ( Main/Merge ); précédent : 00B648; suivant : 00B650A General Joint Additive and Convolutive Bias Compensation Approach Applied to Noisy Lombard Speech Recognition
Auteurs : Mohamed Afify ; Yifan Gong ; Jean-Paul Haton [France]Source :
- IEEE transactions on Speech and Audio processing ; 1998.
English descriptors
- KwdEn :
Abstract
In this paper, a unified approach to the acoustic mismatch problem is proposed. A maximum likelihood state-based additive bias compensation algorithm is developed for the continuous density hidden Markov model (CDHMM). Based on this technique, specific bias models in the mel cepstral and the linear spectral domains are presented. Among these a new polynomial trend bias model in the mel cepstral domain is derived, which proved effective for Lombard speech compensation. In addition, a joint estimation algorithm for additive and convolutive bias compensation is proposed. This algorithm is based on applying the above EM technique in both domains, in conjunction with a parallel model combination (PMC) based transformation. The compensation of the difference coefficients in the proposed framework is also studied.The database c onsists of a 21 confusable word vocabulary uttered by 24 speakers. Three mismatched versions of the database are considered,i.e., Lombar d speech, 15 dB noisy Lombard speech, and 5 dB noisy Lombard speech. The proposed techniques result in 50.9========percnt;, 74.6========percnt;, and 67.3========percnt; reduction in the performance difference between matched and uncompensated word error rates for the three mismatch conditions, respectively. When dynamic coefficients are considered the corresponding reductions are 46.8========percnt;, 72.4========percnt;, and 70.9========percnt;.
Links toward previous steps (curation, corpus...)
- to stream Crin, to step Corpus: 002409
- to stream Crin, to step Curation: 002409
- to stream Crin, to step Checkpoint: 002176
Links to Exploration step
CRIN:afify98bLe document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en" wicri:score="703">A General Joint Additive and Convolutive Bias Compensation Approach Applied to Noisy Lombard Speech Recognition</title>
</titleStmt>
<publicationStmt><idno type="RBID">CRIN:afify98b</idno>
<date when="1998" year="1998">1998</date>
<idno type="wicri:Area/Crin/Corpus">002409</idno>
<idno type="wicri:Area/Crin/Curation">002409</idno>
<idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Curation">002409</idno>
<idno type="wicri:Area/Crin/Checkpoint">002176</idno>
<idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Checkpoint">002176</idno>
<idno type="wicri:Area/Main/Merge">00B649</idno>
</publicationStmt>
<sourceDesc><biblStruct><analytic><title xml:lang="en">A General Joint Additive and Convolutive Bias Compensation Approach Applied to Noisy Lombard Speech Recognition</title>
<author><name sortKey="Afify, Mohamed" sort="Afify, Mohamed" uniqKey="Afify M" first="Mohamed" last="Afify">Mohamed Afify</name>
</author>
<author><name sortKey="Gong, Yifan" sort="Gong, Yifan" uniqKey="Gong Y" first="Yifan" last="Gong">Yifan Gong</name>
</author>
<author><name sortKey="Haton, Jean Paul" sort="Haton, Jean Paul" uniqKey="Haton J" first="Jean-Paul" last="Haton">Jean-Paul Haton</name>
<affiliation><country>France</country>
<placeName><settlement type="city">Nancy</settlement>
<region type="region" nuts="2">Grand Est</region>
<region type="region" nuts="2">Lorraine (région)</region>
</placeName>
<orgName type="laboratoire" n="5">Laboratoire lorrain de recherche en informatique et ses applications</orgName>
<orgName type="university">Université de Lorraine</orgName>
<orgName type="institution">Centre national de la recherche scientifique</orgName>
<orgName type="institution">Institut national de recherche en informatique et en automatique</orgName>
</affiliation>
</author>
</analytic>
<series><title level="j">IEEE transactions on Speech and Audio processing</title>
<imprint><date when="1998" type="published">1998</date>
</imprint>
</series>
</biblStruct>
</sourceDesc>
</fileDesc>
<profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>HMM</term>
<term>Lombard Speech</term>
<term>compensation</term>
<term>noise</term>
</keywords>
</textClass>
</profileDesc>
</teiHeader>
<front><div type="abstract" xml:lang="en" wicri:score="3806">In this paper, a unified approach to the acoustic mismatch problem is proposed. A maximum likelihood state-based additive bias compensation algorithm is developed for the continuous density hidden Markov model (CDHMM). Based on this technique, specific bias models in the mel cepstral and the linear spectral domains are presented. Among these a new polynomial trend bias model in the mel cepstral domain is derived, which proved effective for Lombard speech compensation. In addition, a joint estimation algorithm for additive and convolutive bias compensation is proposed. This algorithm is based on applying the above EM technique in both domains, in conjunction with a parallel model combination (PMC) based transformation. The compensation of the difference coefficients in the proposed framework is also studied.The database c onsists of a 21 confusable word vocabulary uttered by 24 speakers. Three mismatched versions of the database are considered,i.e., Lombar d speech, 15 dB noisy Lombard speech, and 5 dB noisy Lombard speech. The proposed techniques result in 50.9========percnt;, 74.6========percnt;, and 67.3========percnt; reduction in the performance difference between matched and uncompensated word error rates for the three mismatch conditions, respectively. When dynamic coefficients are considered the corresponding reductions are 46.8========percnt;, 72.4========percnt;, and 70.9========percnt;.</div>
</front>
</TEI>
</record>
Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/Main/Merge
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 00B649 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Main/Merge/biblio.hfd -nk 00B649 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien |wiki= Wicri/Lorraine |area= InforLorV4 |flux= Main |étape= Merge |type= RBID |clé= CRIN:afify98b |texte= A General Joint Additive and Convolutive Bias Compensation Approach Applied to Noisy Lombard Speech Recognition }}
![]() | This area was generated with Dilib version V0.6.33. | ![]() |