Frequency and Wavelet Filtering for Robust Speech Recognition
Identifieur interne : 000C27 ( Crin/Checkpoint ); précédent : 000C26; suivant : 000C28Frequency and Wavelet Filtering for Robust Speech Recognition
Auteurs : Murat Deviren ; Khalid DaoudiSource :
English descriptors
Abstract
Mel-frequency cepstral coefficients (MFCC) are the most widely used features in current speech recognition systems. However, they have a poor physical interpretation and they do not lie in the frequency domain. Frequency filtering (FF) is a technique that has been recently developed to design frequency-localized speech features that perform similar to MFCC in terms of recognition performances. Motivated by our desire to build time-frequency speech models, we wanted to use the FF technique as front-end. However, when evaluating FF on the Aurora-3 database we found some discrepancies in the highly mismatch case. This led us to put FF in another perspective : the wavelet transform. By doing so, we were able to explain the discrepancies and to achieve significant improvements in recognition in the highly mismatch case.
Links toward previous steps (curation, corpus...)
Links to Exploration step
CRIN:deviren03aLe document en format XML
<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en" wicri:score="296">Frequency and Wavelet Filtering for Robust Speech Recognition</title>
</titleStmt>
<publicationStmt><idno type="RBID">CRIN:deviren03a</idno>
<date when="2003" year="2003">2003</date>
<idno type="wicri:Area/Crin/Corpus">003C43</idno>
<idno type="wicri:Area/Crin/Curation">003C43</idno>
<idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Curation">003C43</idno>
<idno type="wicri:Area/Crin/Checkpoint">000C27</idno>
<idno type="wicri:explorRef" wicri:stream="Crin" wicri:step="Checkpoint">000C27</idno>
</publicationStmt>
<sourceDesc><biblStruct><analytic><title xml:lang="en">Frequency and Wavelet Filtering for Robust Speech Recognition</title>
<author><name sortKey="Deviren, Murat" sort="Deviren, Murat" uniqKey="Deviren M" first="Murat" last="Deviren">Murat Deviren</name>
</author>
<author><name sortKey="Daoudi, Khalid" sort="Daoudi, Khalid" uniqKey="Daoudi K" first="Khalid" last="Daoudi">Khalid Daoudi</name>
</author>
</analytic>
</biblStruct>
</sourceDesc>
</fileDesc>
<profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>frequency filtering</term>
<term>noise robustness</term>
<term>speech recognition</term>
<term>wavelets</term>
</keywords>
</textClass>
</profileDesc>
</teiHeader>
<front><div type="abstract" xml:lang="en" wicri:score="1836">Mel-frequency cepstral coefficients (MFCC) are the most widely used features in current speech recognition systems. However, they have a poor physical interpretation and they do not lie in the frequency domain. Frequency filtering (FF) is a technique that has been recently developed to design frequency-localized speech features that perform similar to MFCC in terms of recognition performances. Motivated by our desire to build time-frequency speech models, we wanted to use the FF technique as front-end. However, when evaluating FF on the Aurora-3 database we found some discrepancies in the highly mismatch case. This led us to put FF in another perspective : the wavelet transform. By doing so, we were able to explain the discrepancies and to achieve significant improvements in recognition in the highly mismatch case.</div>
</front>
</TEI>
<BibTex type="inproceedings"><ref>deviren03a</ref>
<crinnumber>A03-R-529</crinnumber>
<category>3</category>
<equipe>PAROLE</equipe>
<author><e>Deviren, Murat</e>
<e>Daoudi, Khalid</e>
</author>
<title>Frequency and Wavelet Filtering for Robust Speech Recognition</title>
<booktitle>{Artificial Neural Networks and Neural Information Processing - Joint International Conference ICANN/ICONIP2003 , Istanbul, Turquie}</booktitle>
<year>2003</year>
<editor>Kaynak, Okyay and Alpaydin Ethem and Oja, Erkki and Xu, Lei</editor>
<volume>2714</volume>
<series>Lectures Notes in Computer Science</series>
<pages>452-462</pages>
<month>Jun</month>
<publisher>Springer-Verlag</publisher>
<keywords><e>speech recognition</e>
<e>noise robustness</e>
<e>frequency filtering</e>
<e>wavelets</e>
</keywords>
<abstract>Mel-frequency cepstral coefficients (MFCC) are the most widely used features in current speech recognition systems. However, they have a poor physical interpretation and they do not lie in the frequency domain. Frequency filtering (FF) is a technique that has been recently developed to design frequency-localized speech features that perform similar to MFCC in terms of recognition performances. Motivated by our desire to build time-frequency speech models, we wanted to use the FF technique as front-end. However, when evaluating FF on the Aurora-3 database we found some discrepancies in the highly mismatch case. This led us to put FF in another perspective : the wavelet transform. By doing so, we were able to explain the discrepancies and to achieve significant improvements in recognition in the highly mismatch case.</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 000C27 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Crin/Checkpoint/biblio.hfd -nk 000C27 | 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:deviren03a |texte= Frequency and Wavelet Filtering for Robust Speech Recognition }}
![]() | This area was generated with Dilib version V0.6.33. | ![]() |