Une nouvelle architecture de compensation du bruit pour la reconnaissance robuste de la parole
Identifieur interne : 006804 ( Main/Exploration ); précédent : 006803; suivant : 006805Une nouvelle architecture de compensation du bruit pour la reconnaissance robuste de la parole
Auteurs : Khalid Daoudi ; Murat DevirenSource :
English descriptors
- KwdEn :
Abstract
We present a novel noise compensation architecture which makes no assumptions on how the noise sources alter the speech data and which do not rely on clean speech models. Rather, this new architecture makes the (realistic) assumption that speech databases recorded under different background noise conditions are available. Its main principle is to process individually each database and to construct a parametric representation which describes the variation of acoustic models w.r.t. noise models. This representation is then used during recognition to estimate the acoustic models in the new environment. We evaluate the performance of this new compensation scheme on a connected digits recognition task and show that it can perform significantly better than multi-conditions training, which is the most widely used technique in these kind of scenarios.
Affiliations:
Links toward previous steps (curation, corpus...)
- to stream Crin, to step Corpus: 003D88
- to stream Crin, to step Curation: 003D88
- to stream Crin, to step Checkpoint: 000632
- to stream Main, to step Merge: 006B07
- to stream Main, to step Curation: 006804
Le document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en" wicri:score="69">Une nouvelle architecture de compensation du bruit pour la reconnaissance robuste de la parole</title></titleStmt><publicationStmt><idno type="RBID">CRIN:daoudi04a</idno><date when="2004" year="2004">2004</date><idno type="wicri:Area/Crin/Corpus">003D88</idno><idno type="wicri:Area/Crin/Curation">003D88</idno><idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Curation">003D88</idno><idno type="wicri:Area/Crin/Checkpoint">000632</idno><idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Checkpoint">000632</idno><idno type="wicri:Area/Main/Merge">006B07</idno><idno type="wicri:Area/Main/Curation">006804</idno><idno type="wicri:Area/Main/Exploration">006804</idno></publicationStmt><sourceDesc><biblStruct><analytic><title xml:lang="en">Une nouvelle architecture de compensation du bruit pour la reconnaissance robuste de la parole</title><author><name sortKey="Daoudi, Khalid" sort="Daoudi, Khalid" uniqKey="Daoudi K" first="Khalid" last="Daoudi">Khalid Daoudi</name></author><author><name sortKey="Deviren, Murat" sort="Deviren, Murat" uniqKey="Deviren M" first="Murat" last="Deviren">Murat Deviren</name></author></analytic></biblStruct></sourceDesc></fileDesc><profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>noise robustness</term><term>speech recognition</term></keywords></textClass></profileDesc></teiHeader><front><div type="abstract" xml:lang="en" wicri:score="2655">We present a novel noise compensation architecture which makes no assumptions on how the noise sources alter the speech data and which do not rely on clean speech models. Rather, this new architecture makes the (realistic) assumption that speech databases recorded under different background noise conditions are available. Its main principle is to process individually each database and to construct a parametric representation which describes the variation of acoustic models w.r.t. noise models. This representation is then used during recognition to estimate the acoustic models in the new environment. We evaluate the performance of this new compensation scheme on a connected digits recognition task and show that it can perform significantly better than multi-conditions training, which is the most widely used technique in these kind of scenarios.</div></front></TEI><affiliations><list></list><tree><noCountry><name sortKey="Daoudi, Khalid" sort="Daoudi, Khalid" uniqKey="Daoudi K" first="Khalid" last="Daoudi">Khalid Daoudi</name><name sortKey="Deviren, Murat" sort="Deviren, Murat" uniqKey="Deviren M" first="Murat" last="Deviren">Murat Deviren</name></noCountry></tree></affiliations></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/Main/Exploration
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 006804 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd -nk 006804 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Wicri/Lorraine
|area= InforLorV4
|flux= Main
|étape= Exploration
|type= RBID
|clé= CRIN:daoudi04a
|texte= Une nouvelle architecture de compensation du bruit pour la reconnaissance robuste de la parole
}}
| This area was generated with Dilib version V0.6.33. |  |