The Importance of Segmentation Probability in Segment Based Speech Recognizers
Identifieur interne : 001E88 ( Crin/Curation ); précédent : 001E87; suivant : 001E89The Importance of Segmentation Probability in Segment Based Speech Recognizers
Auteurs : Jan Verhasselt ; Irina Illina ; Jean-Pierre Martens ; Yifan Gong ; Jean-Paul HatonSource :
English descriptors
Abstract
In segment based recognizers, variable length speech segments are mapped to the basic speech units (phones, diphones, ...). In this paper, we address the acoustical modeling of these basic units in the framework of segmental posterior distribution models (SPDM). The joint posterior probability of a unit sequence \underline{u} and a segmentation \underline{s}, Pr(\underline{u},\underline{s}|\underline{\bf x}) can be written as the product of the segmentation probability Pr(\underline{s}|\underline{bf x}) and the unit classification probability Pr(\underline{u}|\underline{s},\underline{\bf x}), where \underline{\bf x} is the sequence of acoustic observation parameter vectors. In particular, we point out the role of the segmentation probability and demonstrate that it does improve the recognition accuracy. We present evidence for this in two different tasks (speaker dependent continuous word recognition in French and speaker independent phone recognition in American English) in combination with two different unit classification models.
Links toward previous steps (curation, corpus...)
- to stream Crin, to step Corpus: Pour aller vers cette notice dans l'étape Curation :001E88
Links to Exploration step
CRIN:verhasselt97bLe document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en" wicri:score="492">The Importance of Segmentation Probability in Segment Based Speech Recognizers</title></titleStmt><publicationStmt><idno type="RBID">CRIN:verhasselt97b</idno><date when="1997" year="1997">1997</date><idno type="wicri:Area/Crin/Corpus">001E88</idno><idno type="wicri:Area/Crin/Curation">001E88</idno><idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Curation">001E88</idno></publicationStmt><sourceDesc><biblStruct><analytic><title xml:lang="en">The Importance of Segmentation Probability in Segment Based Speech Recognizers</title><author><name sortKey="Verhasselt, Jan" sort="Verhasselt, Jan" uniqKey="Verhasselt J" first="Jan" last="Verhasselt">Jan Verhasselt</name></author><author><name sortKey="Illina, Irina" sort="Illina, Irina" uniqKey="Illina I" first="Irina" last="Illina">Irina Illina</name></author><author><name sortKey="Martens, Jean Pierre" sort="Martens, Jean Pierre" uniqKey="Martens J" first="Jean-Pierre" last="Martens">Jean-Pierre Martens</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></author></analytic></biblStruct></sourceDesc></fileDesc><profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>segment-based recognizer</term><term>speech recognition</term></keywords></textClass></profileDesc></teiHeader><front><div type="abstract" xml:lang="en" wicri:score="2697">In segment based recognizers, variable length speech segments are mapped to the basic speech units (phones, diphones, ...). In this paper, we address the acoustical modeling of these basic units in the framework of segmental posterior distribution models (SPDM). The joint posterior probability of a unit sequence \underline{u} and a segmentation \underline{s}, Pr(\underline{u},\underline{s}|\underline{\bf x}) can be written as the product of the segmentation probability Pr(\underline{s}|\underline{bf x}) and the unit classification probability Pr(\underline{u}|\underline{s},\underline{\bf x}), where \underline{\bf x} is the sequence of acoustic observation parameter vectors. In particular, we point out the role of the segmentation probability and demonstrate that it does improve the recognition accuracy. We present evidence for this in two different tasks (speaker dependent continuous word recognition in French and speaker independent phone recognition in American English) in combination with two different unit classification models.</div></front></TEI><BibTex type="inproceedings"><ref>verhasselt97b</ref><crinnumber>97-R-002</crinnumber><category>3</category><equipe>RFIA</equipe><author><e>Verhasselt, Jan</e><e>Illina, Irina</e><e>Martens, Jean-Pierre</e><e>Gong, Yifan</e><e>Haton, Jean-Paul</e></author><title>The Importance of Segmentation Probability in Segment Based Speech Recognizers</title><booktitle>{International Conference on Acoustics, Speech, and Signal Processing - ICASSP'97, Munich, Germany}</booktitle><year>1997</year><volume>2</volume><pages>1407-1410</pages><month>apr</month><publisher>IEEE Computer society Press</publisher><keywords><e>speech recognition</e><e>segment-based recognizer</e></keywords><abstract>In segment based recognizers, variable length speech segments are mapped to the basic speech units (phones, diphones, ...). In this paper, we address the acoustical modeling of these basic units in the framework of segmental posterior distribution models (SPDM). The joint posterior probability of a unit sequence \underline{u} and a segmentation \underline{s}, Pr(\underline{u},\underline{s}|\underline{\bf x}) can be written as the product of the segmentation probability Pr(\underline{s}|\underline{bf x}) and the unit classification probability Pr(\underline{u}|\underline{s},\underline{\bf x}), where \underline{\bf x} is the sequence of acoustic observation parameter vectors. In particular, we point out the role of the segmentation probability and demonstrate that it does improve the recognition accuracy. We present evidence for this in two different tasks (speaker dependent continuous word recognition in French and speaker independent phone recognition in American English) in combination with two different unit classification models.</abstract></BibTex></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/Crin/Curation
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 001E88 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Crin/Curation/biblio.hfd -nk 001E88 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Wicri/Lorraine
|area= InforLorV4
|flux= Crin
|étape= Curation
|type= RBID
|clé= CRIN:verhasselt97b
|texte= The Importance of Segmentation Probability in Segment Based Speech Recognizers
}}
| This area was generated with Dilib version V0.6.33. |  |