Détection automatique de sons bien réalisés
Identifieur interne : 000604 ( Crin/Checkpoint ); précédent : 000603; suivant : 000605Détection automatique de sons bien réalisés
Auteurs : Yves Laprie ; Safaa Jarifi ; Anne Bonneau ; Dominique FohrSource :
English descriptors
- KwdEn :
Abstract
Given a phonetic context, sounds can be uttered with more or less salient acoustic cues depending on the speech style and prosody. In a previous work we studied strong acoustic cues of unvoiced stops that enable a very reliable identification of stops. In this paper we use this background idea again with a view of exploiting well realized sounds to enhance speech intelligibility within the framework of language learning. We thus designed an elitist learning of HMM that make very reliable phone models emerge. The learning is iterated by feeding phones identified correctly at the previous iteration into the learning algorithm. In this way models specialize to represent well realized sounds. Experiments were carried out on the BREF 80 corpus by constructing well realized phone models for unvoiced stops. They show that these contextual models triggered off in 60% of stops occurrences with an extremely low confusion rate.
Links toward previous steps (curation, corpus...)
Links to Exploration step
CRIN:laprie04aLe document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="fr" wicri:score="-100">Détection automatique de sons bien réalisés</title>
</titleStmt>
<publicationStmt><idno type="RBID">CRIN:laprie04a</idno>
<date when="2004" year="2004">2004</date>
<idno type="wicri:Area/Crin/Corpus">003F61</idno>
<idno type="wicri:Area/Crin/Curation">003F61</idno>
<idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Curation">003F61</idno>
<idno type="wicri:Area/Crin/Checkpoint">000604</idno>
<idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Checkpoint">000604</idno>
</publicationStmt>
<sourceDesc><biblStruct><analytic><title xml:lang="fr">Détection automatique de sons bien réalisés</title>
<author><name sortKey="Laprie, Yves" sort="Laprie, Yves" uniqKey="Laprie Y" first="Yves" last="Laprie">Yves Laprie</name>
</author>
<author><name sortKey="Jarifi, Safaa" sort="Jarifi, Safaa" uniqKey="Jarifi S" first="Safaa" last="Jarifi">Safaa Jarifi</name>
</author>
<author><name sortKey="Bonneau, Anne" sort="Bonneau, Anne" uniqKey="Bonneau A" first="Anne" last="Bonneau">Anne Bonneau</name>
</author>
<author><name sortKey="Fohr, Dominique" sort="Fohr, Dominique" uniqKey="Fohr D" first="Dominique" last="Fohr">Dominique Fohr</name>
</author>
</analytic>
</biblStruct>
</sourceDesc>
</fileDesc>
<profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>acoustic cues</term>
<term>automatic recognition</term>
<term>markov models</term>
<term>sounds</term>
</keywords>
</textClass>
</profileDesc>
</teiHeader>
<front><div type="abstract" xml:lang="en" wicri:score="2776">Given a phonetic context, sounds can be uttered with more or less salient acoustic cues depending on the speech style and prosody. In a previous work we studied strong acoustic cues of unvoiced stops that enable a very reliable identification of stops. In this paper we use this background idea again with a view of exploiting well realized sounds to enhance speech intelligibility within the framework of language learning. We thus designed an elitist learning of HMM that make very reliable phone models emerge. The learning is iterated by feeding phones identified correctly at the previous iteration into the learning algorithm. In this way models specialize to represent well realized sounds. Experiments were carried out on the BREF 80 corpus by constructing well realized phone models for unvoiced stops. They show that these contextual models triggered off in 60% of stops occurrences with an extremely low confusion rate.</div>
</front>
</TEI>
<BibTex type="inproceedings"><ref>laprie04a</ref>
<crinnumber>A04-R-284</crinnumber>
<category>3</category>
<equipe>PAROLE</equipe>
<author><e>Laprie, Yves</e>
<e>Jarifi, Safaa</e>
<e>Bonneau, Anne</e>
<e>Fohr, Dominique</e>
</author>
<title>Détection automatique de sons bien réalisés</title>
<booktitle>{Actes des XXVes Journées d'Étude sur la Parole - JEP'2004, Fès, Maroc}</booktitle>
<year>2004</year>
<month>Apr</month>
<url>http://www.loria.fr/publications/2004/A04-R-284/A04-R-284.ps</url>
<keywords><e>automatic recognition</e>
<e>sounds</e>
<e>markov models</e>
<e>acoustic cues</e>
</keywords>
<abstract>Given a phonetic context, sounds can be uttered with more or less salient acoustic cues depending on the speech style and prosody. In a previous work we studied strong acoustic cues of unvoiced stops that enable a very reliable identification of stops. In this paper we use this background idea again with a view of exploiting well realized sounds to enhance speech intelligibility within the framework of language learning. We thus designed an elitist learning of HMM that make very reliable phone models emerge. The learning is iterated by feeding phones identified correctly at the previous iteration into the learning algorithm. In this way models specialize to represent well realized sounds. Experiments were carried out on the BREF 80 corpus by constructing well realized phone models for unvoiced stops. They show that these contextual models triggered off in 60% of stops occurrences with an extremely low confusion rate.</abstract>
</BibTex>
</record>
Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/Crin/Checkpoint
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000604 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Crin/Checkpoint/biblio.hfd -nk 000604 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien |wiki= Wicri/Lorraine |area= InforLorV4 |flux= Crin |étape= Checkpoint |type= RBID |clé= CRIN:laprie04a |texte= Détection automatique de sons bien réalisés }}
![]() | This area was generated with Dilib version V0.6.33. | ![]() |