InforLorV4, PascalFrancis, Corpus, bibRecord, 000627

On the use of high order derivatives for high performance alphabet recognition

Identifieur interne : 000627 ( PascalFrancis/Corpus ); précédent : 000626; suivant : 000628

On the use of high order derivatives for high performance alphabet recognition

Auteurs : Joseph Di Martino

Source :

Proceedings of the ... IEEE International Conference on Acoustics, Speech and Signal Processing [ 1520-6149 ] ; 2002.

RBID : Pascal:04-0510947

Descripteurs français

Pascal (Inist)
- Haute performance, Reconnaissance caractère, Reconnaissance automatique, Reconnaissance parole, Analyse spectre, Analyse cepstrale, Base donnée, Modèle Markov variable cachée, Précision, Intervalle confiance, Segmentation, Algorithme, Implémentation, Reconnaissance forme, Traitement parole, Analyse signal, Approche probabiliste, Extraction caractéristique, Traitement signal.

English descriptors

KwdEn :
- Accuracy, Algorithm, Automatic recognition, Cepstral analysis, Character recognition, Confidence interval, Database, Feature extraction, Hidden Markov models, High performance, Implementation, Pattern recognition, Probabilistic approach, Segmentation, Signal analysis, Signal processing, Spectrum analysis, Speech processing, Speech recognition.

Abstract

In this paper I propose new feature vectors for automatic speech recognition. They are based on Mel-cepstrum vectors augmented by derivatives. In the literature, many systems using just two derivatives-delta and delta delta- are described. But none explores the use of higher order derivatives. This paper presents alphabet recognition results on the Isolet database, using feature vectors containing up to the fifth-order derivatives. For this paper I did not use the HTK toolkit proposed by Cambridge University. I developed my own HMM system. I show that with vectors incorporating all the derivatives up to the fifth one, 97.54% mean recognition accuracy was achieved, result which is comparable to the best published one on this database (97.6%), if the recognition accuracy confidence interval concerning this task (approximately 0.3%) is taken into account. It is important to note that this result was obtained without segmenting the speech files by an endpoint detection algorithm. This is an unfavourable experimental condition compared to previous published research works. As a consequence, my system is one of the most powerful systems ever implemented for alphabet recognition.

Notice en format standard (ISO 2709)

Pour connaître la documentation sur le format Inist Standard.

A01	`01`	`1`		`@0 1520-6149`
A08	`01`	`1`	`ENG`	`@1 On the use of high order derivatives for high performance alphabet recognition`
A09	`01`	`1`	`ENG`	@1 2002 IEEE international conference on acoustics, speech, and signal processing : Orlando FL, 13-17 May 2002. Volume I: Speech processing, neural networks for signal processing. Volume II: Signal processing theory and methods, audio and electro-acoustics, multimedia signal processing. Volume III: Signal processing for communications, sensor array and multichannel signal processing, design and implementation of signal processing systems. Volume IV: Image and multidimensional signal processing, industry technology tracks, special sessions
A11	`01`	`1`		`@1 DI MARTINO (Joseph)`
A14	`01`			`@1 LORIA, B.P 239 @2 Vandœuvre-lès-Nancy 54506 @3 FRA @Z 1 aut.`
A18	`01`	`1`		`@1 IEEE Signal Processing Society @3 USA @9 patr.`
A20				`@2 vol I, 953-956`
A21				`@1 2002`
A23	`01`			`@0 ENG`
A26	`01`			`@0 0-7803-7402-9`
A43	`01`			`@1 INIST @2 Y 38009 @5 354000117914991195`
A44				`@0 0000 @1 © 2004 INIST-CNRS. All rights reserved.`
A45				`@0 10 ref.`
A47	`01`	`1`		`@0 04-0510947`
A60				`@1 P @2 C`
A61				`@0 A`
A64	`01`	`1`		`@0 Proceedings of the ... IEEE International Conference on Acoustics, Speech and Signal Processing`
A66	`01`			`@0 USA`
C01	`01`		`ENG`	@0 In this paper I propose new feature vectors for automatic speech recognition. They are based on Mel-cepstrum vectors augmented by derivatives. In the literature, many systems using just two derivatives-delta and delta delta- are described. But none explores the use of higher order derivatives. This paper presents alphabet recognition results on the Isolet database, using feature vectors containing up to the fifth-order derivatives. For this paper I did not use the HTK toolkit proposed by Cambridge University. I developed my own HMM system. I show that with vectors incorporating all the derivatives up to the fifth one, 97.54% mean recognition accuracy was achieved, result which is comparable to the best published one on this database (97.6%), if the recognition accuracy confidence interval concerning this task (approximately 0.3%) is taken into account. It is important to note that this result was obtained without segmenting the speech files by an endpoint detection algorithm. This is an unfavourable experimental condition compared to previous published research works. As a consequence, my system is one of the most powerful systems ever implemented for alphabet recognition.
C02	`01`	`X`		`@0 001D04A05A`
C02	`02`	`X`		`@0 001D04A05B`
C02	`03`	`X`		`@0 001D04A04A1`
C03	`01`	`X`	`FRE`	`@0 Haute performance @5 01`
C03	`01`	`X`	`ENG`	`@0 High performance @5 01`
C03	`01`	`X`	`SPA`	`@0 Alto rendimiento @5 01`
C03	`02`	`X`	`FRE`	`@0 Reconnaissance caractère @5 02`
C03	`02`	`X`	`ENG`	`@0 Character recognition @5 02`
C03	`02`	`X`	`SPA`	`@0 Reconocimiento carácter @5 02`
C03	`03`	`X`	`FRE`	`@0 Reconnaissance automatique @5 03`
C03	`03`	`X`	`ENG`	`@0 Automatic recognition @5 03`
C03	`03`	`X`	`SPA`	`@0 Reconocimiento automático @5 03`
C03	`04`	`X`	`FRE`	`@0 Reconnaissance parole @5 04`
C03	`04`	`X`	`ENG`	`@0 Speech recognition @5 04`
C03	`04`	`X`	`SPA`	`@0 Reconocimiento voz @5 04`
C03	`05`	`X`	`FRE`	`@0 Analyse spectre @5 05`
C03	`05`	`X`	`ENG`	`@0 Spectrum analysis @5 05`
C03	`05`	`X`	`SPA`	`@0 Análisis espectro @5 05`
C03	`06`	`3`	`FRE`	`@0 Analyse cepstrale @5 06`
C03	`06`	`3`	`ENG`	`@0 Cepstral analysis @5 06`
C03	`07`	`X`	`FRE`	`@0 Base donnée @5 07`
C03	`07`	`X`	`ENG`	`@0 Database @5 07`
C03	`07`	`X`	`SPA`	`@0 Base dato @5 07`
C03	`08`	`3`	`FRE`	`@0 Modèle Markov variable cachée @5 08`
C03	`08`	`3`	`ENG`	`@0 Hidden Markov models @5 08`
C03	`09`	`X`	`FRE`	`@0 Précision @5 09`
C03	`09`	`X`	`ENG`	`@0 Accuracy @5 09`
C03	`09`	`X`	`SPA`	`@0 Precisión @5 09`
C03	`10`	`X`	`FRE`	`@0 Intervalle confiance @5 10`
C03	`10`	`X`	`ENG`	`@0 Confidence interval @5 10`
C03	`10`	`X`	`SPA`	`@0 Intervalo confianza @5 10`
C03	`11`	`X`	`FRE`	`@0 Segmentation @5 11`
C03	`11`	`X`	`ENG`	`@0 Segmentation @5 11`
C03	`11`	`X`	`SPA`	`@0 Segmentación @5 11`
C03	`12`	`X`	`FRE`	`@0 Algorithme @5 12`
C03	`12`	`X`	`ENG`	`@0 Algorithm @5 12`
C03	`12`	`X`	`SPA`	`@0 Algoritmo @5 12`
C03	`13`	`X`	`FRE`	`@0 Implémentation @5 13`
C03	`13`	`X`	`ENG`	`@0 Implementation @5 13`
C03	`13`	`X`	`SPA`	`@0 Implementación @5 13`
C03	`14`	`X`	`FRE`	`@0 Reconnaissance forme @5 14`
C03	`14`	`X`	`ENG`	`@0 Pattern recognition @5 14`
C03	`14`	`X`	`SPA`	`@0 Reconocimiento patrón @5 14`
C03	`15`	`X`	`FRE`	`@0 Traitement parole @5 15`
C03	`15`	`X`	`ENG`	`@0 Speech processing @5 15`
C03	`15`	`X`	`SPA`	`@0 Tratamiento palabra @5 15`
C03	`16`	`X`	`FRE`	`@0 Analyse signal @5 16`
C03	`16`	`X`	`ENG`	`@0 Signal analysis @5 16`
C03	`16`	`X`	`SPA`	`@0 Análisis de señal @5 16`
C03	`17`	`X`	`FRE`	`@0 Approche probabiliste @5 17`
C03	`17`	`X`	`ENG`	`@0 Probabilistic approach @5 17`
C03	`17`	`X`	`SPA`	`@0 Enfoque probabilista @5 17`
C03	`18`	`3`	`FRE`	`@0 Extraction caractéristique @5 18`
C03	`18`	`3`	`ENG`	`@0 Feature extraction @5 18`
C03	`19`	`X`	`FRE`	`@0 Traitement signal @5 19`
C03	`19`	`X`	`ENG`	`@0 Signal processing @5 19`
C03	`19`	`X`	`SPA`	`@0 Procesamiento señal @5 19`
N21				`@1 285`
N44	`01`			`@1 OTO`
N82				`@1 OTO`

A30	`01`	`1`	`ENG`	`@1 International conference on acoustics, speech, and signal processing @3 Orlando FL USA @4 2002-05-13`

Format Inist (serveur)

NO :	PASCAL 04-0510947 INIST
ET :	On the use of high order derivatives for high performance alphabet recognition
AU :	DI MARTINO (Joseph)
AF :	LORIA, B.P 239/Vandœuvre-lès-Nancy 54506/France (1 aut.)
DT :	Publication en série; Congrès; Niveau analytique
SO :	Proceedings of the ... IEEE International Conference on Acoustics, Speech and Signal Processing; ISSN 1520-6149; Etats-Unis; Da. 2002; vol I, 953-956; Bibl. 10 ref.
LA :	Anglais
EA :	In this paper I propose new feature vectors for automatic speech recognition. They are based on Mel-cepstrum vectors augmented by derivatives. In the literature, many systems using just two derivatives-delta and delta delta- are described. But none explores the use of higher order derivatives. This paper presents alphabet recognition results on the Isolet database, using feature vectors containing up to the fifth-order derivatives. For this paper I did not use the HTK toolkit proposed by Cambridge University. I developed my own HMM system. I show that with vectors incorporating all the derivatives up to the fifth one, 97.54% mean recognition accuracy was achieved, result which is comparable to the best published one on this database (97.6%), if the recognition accuracy confidence interval concerning this task (approximately 0.3%) is taken into account. It is important to note that this result was obtained without segmenting the speech files by an endpoint detection algorithm. This is an unfavourable experimental condition compared to previous published research works. As a consequence, my system is one of the most powerful systems ever implemented for alphabet recognition.
CC :	001D04A05A; 001D04A05B; 001D04A04A1
FD :	Haute performance; Reconnaissance caractère; Reconnaissance automatique; Reconnaissance parole; Analyse spectre; Analyse cepstrale; Base donnée; Modèle Markov variable cachée; Précision; Intervalle confiance; Segmentation; Algorithme; Implémentation; Reconnaissance forme; Traitement parole; Analyse signal; Approche probabiliste; Extraction caractéristique; Traitement signal
ED :	High performance; Character recognition; Automatic recognition; Speech recognition; Spectrum analysis; Cepstral analysis; Database; Hidden Markov models; Accuracy; Confidence interval; Segmentation; Algorithm; Implementation; Pattern recognition; Speech processing; Signal analysis; Probabilistic approach; Feature extraction; Signal processing
SD :	Alto rendimiento; Reconocimiento carácter; Reconocimiento automático; Reconocimiento voz; Análisis espectro; Base dato; Precisión; Intervalo confianza; Segmentación; Algoritmo; Implementación; Reconocimiento patrón; Tratamiento palabra; Análisis de señal; Enfoque probabilista; Procesamiento señal
LO :	INIST-Y 38009.354000117914991195
ID :	04-0510947

Links to Exploration step

Pascal:04-0510947

Le document en format XML

<record><TEI><teiHeader><fileDesc><titleStmt><title xml:lang="en" level="a">On the use of high order derivatives for high performance alphabet recognition</title>
<author><name sortKey="Di Martino, Joseph" sort="Di Martino, Joseph" uniqKey="Di Martino J" first="Joseph" last="Di Martino">Joseph Di Martino</name>
<affiliation><inist:fA14 i1="01"><s1>LORIA, B.P 239</s1>
<s2>Vandœuvre-lès-Nancy 54506</s2>
<s3>FRA</s3>
<sZ>1 aut.</sZ>
</inist:fA14>
</affiliation>
</author>
</titleStmt>
<publicationStmt><idno type="wicri:source">INIST</idno>
<idno type="inist">04-0510947</idno>
<date when="2002">2002</date>
<idno type="stanalyst">PASCAL 04-0510947 INIST</idno>
<idno type="RBID">Pascal:04-0510947</idno>
<idno type="wicri:Area/PascalFrancis/Corpus">000627</idno>
</publicationStmt>
<sourceDesc><biblStruct><analytic><title xml:lang="en" level="a">On the use of high order derivatives for high performance alphabet recognition</title>
<author><name sortKey="Di Martino, Joseph" sort="Di Martino, Joseph" uniqKey="Di Martino J" first="Joseph" last="Di Martino">Joseph Di Martino</name>
<affiliation><inist:fA14 i1="01"><s1>LORIA, B.P 239</s1>
<s2>Vandœuvre-lès-Nancy 54506</s2>
<s3>FRA</s3>
<sZ>1 aut.</sZ>
</inist:fA14>
</affiliation>
</author>
</analytic>
<series><title level="j" type="main">Proceedings of the ... IEEE International Conference on Acoustics, Speech and Signal Processing</title>
<idno type="ISSN">1520-6149</idno>
<imprint><date when="2002">2002</date>
</imprint>
</series>
</biblStruct>
</sourceDesc>
<seriesStmt><title level="j" type="main">Proceedings of the ... IEEE International Conference on Acoustics, Speech and Signal Processing</title>
<idno type="ISSN">1520-6149</idno>
</seriesStmt>
</fileDesc>
<profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>Accuracy</term>
<term>Algorithm</term>
<term>Automatic recognition</term>
<term>Cepstral analysis</term>
<term>Character recognition</term>
<term>Confidence interval</term>
<term>Database</term>
<term>Feature extraction</term>
<term>Hidden Markov models</term>
<term>High performance</term>
<term>Implementation</term>
<term>Pattern recognition</term>
<term>Probabilistic approach</term>
<term>Segmentation</term>
<term>Signal analysis</term>
<term>Signal processing</term>
<term>Spectrum analysis</term>
<term>Speech processing</term>
<term>Speech recognition</term>
</keywords>
<keywords scheme="Pascal" xml:lang="fr"><term>Haute performance</term>
<term>Reconnaissance caractère</term>
<term>Reconnaissance automatique</term>
<term>Reconnaissance parole</term>
<term>Analyse spectre</term>
<term>Analyse cepstrale</term>
<term>Base donnée</term>
<term>Modèle Markov variable cachée</term>
<term>Précision</term>
<term>Intervalle confiance</term>
<term>Segmentation</term>
<term>Algorithme</term>
<term>Implémentation</term>
<term>Reconnaissance forme</term>
<term>Traitement parole</term>
<term>Analyse signal</term>
<term>Approche probabiliste</term>
<term>Extraction caractéristique</term>
<term>Traitement signal</term>
</keywords>
</textClass>
</profileDesc>
</teiHeader>
<front><div type="abstract" xml:lang="en">In this paper I propose new feature vectors for automatic speech recognition. They are based on Mel-cepstrum vectors augmented by derivatives. In the literature, many systems using just two derivatives-delta and delta delta- are described. But none explores the use of higher order derivatives. This paper presents alphabet recognition results on the Isolet database, using feature vectors containing up to the fifth-order derivatives. For this paper I did not use the HTK toolkit proposed by Cambridge University. I developed my own HMM system. I show that with vectors incorporating all the derivatives up to the fifth one, 97.54% mean recognition accuracy was achieved, result which is comparable to the best published one on this database (97.6%), if the recognition accuracy confidence interval concerning this task (approximately 0.3%) is taken into account. It is important to note that this result was obtained without segmenting the speech files by an endpoint detection algorithm. This is an unfavourable experimental condition compared to previous published research works. As a consequence, my system is one of the most powerful systems ever implemented for alphabet recognition.</div>
</front>
</TEI>
<inist><standard h6="B"><pA><fA01 i1="01" i2="1"><s0>1520-6149</s0>
</fA01>
<fA08 i1="01" i2="1" l="ENG"><s1>On the use of high order derivatives for high performance alphabet recognition</s1>
</fA08>
<fA09 i1="01" i2="1" l="ENG"><s1>2002 IEEE international conference on acoustics, speech, and signal processing : Orlando FL, 13-17 May 2002. Volume I: Speech processing, neural networks for signal processing. Volume II: Signal processing theory and methods, audio and electro-acoustics, multimedia signal processing. Volume III: Signal processing for communications, sensor array and multichannel signal processing, design and implementation of signal processing systems. Volume IV: Image and multidimensional signal processing, industry technology tracks, special sessions</s1>
</fA09>
<fA11 i1="01" i2="1"><s1>DI MARTINO (Joseph)</s1>
</fA11>
<fA14 i1="01"><s1>LORIA, B.P 239</s1>
<s2>Vandœuvre-lès-Nancy 54506</s2>
<s3>FRA</s3>
<sZ>1 aut.</sZ>
</fA14>
<fA18 i1="01" i2="1"><s1>IEEE Signal Processing Society</s1>
<s3>USA</s3>
<s9>patr.</s9>
</fA18>
<fA20><s2>vol I, 953-956</s2>
</fA20>
<fA21><s1>2002</s1>
</fA21>
<fA23 i1="01"><s0>ENG</s0>
</fA23>
<fA26 i1="01"><s0>0-7803-7402-9</s0>
</fA26>
<fA43 i1="01"><s1>INIST</s1>
<s2>Y 38009</s2>
<s5>354000117914991195</s5>
</fA43>
<fA44><s0>0000</s0>
<s1>© 2004 INIST-CNRS. All rights reserved.</s1>
</fA44>
<fA45><s0>10 ref.</s0>
</fA45>
<fA47 i1="01" i2="1"><s0>04-0510947</s0>
</fA47>
<fA60><s1>P</s1>
<s2>C</s2>
</fA60>
<fA61><s0>A</s0>
</fA61>
<fA64 i1="01" i2="1"><s0>Proceedings of the ... IEEE International Conference on Acoustics, Speech and Signal Processing</s0>
</fA64>
<fA66 i1="01"><s0>USA</s0>
</fA66>
<fC01 i1="01" l="ENG"><s0>In this paper I propose new feature vectors for automatic speech recognition. They are based on Mel-cepstrum vectors augmented by derivatives. In the literature, many systems using just two derivatives-delta and delta delta- are described. But none explores the use of higher order derivatives. This paper presents alphabet recognition results on the Isolet database, using feature vectors containing up to the fifth-order derivatives. For this paper I did not use the HTK toolkit proposed by Cambridge University. I developed my own HMM system. I show that with vectors incorporating all the derivatives up to the fifth one, 97.54% mean recognition accuracy was achieved, result which is comparable to the best published one on this database (97.6%), if the recognition accuracy confidence interval concerning this task (approximately 0.3%) is taken into account. It is important to note that this result was obtained without segmenting the speech files by an endpoint detection algorithm. This is an unfavourable experimental condition compared to previous published research works. As a consequence, my system is one of the most powerful systems ever implemented for alphabet recognition.</s0>
</fC01>
<fC02 i1="01" i2="X"><s0>001D04A05A</s0>
</fC02>
<fC02 i1="02" i2="X"><s0>001D04A05B</s0>
</fC02>
<fC02 i1="03" i2="X"><s0>001D04A04A1</s0>
</fC02>
<fC03 i1="01" i2="X" l="FRE"><s0>Haute performance</s0>
<s5>01</s5>
</fC03>
<fC03 i1="01" i2="X" l="ENG"><s0>High performance</s0>
<s5>01</s5>
</fC03>
<fC03 i1="01" i2="X" l="SPA"><s0>Alto rendimiento</s0>
<s5>01</s5>
</fC03>
<fC03 i1="02" i2="X" l="FRE"><s0>Reconnaissance caractère</s0>
<s5>02</s5>
</fC03>
<fC03 i1="02" i2="X" l="ENG"><s0>Character recognition</s0>
<s5>02</s5>
</fC03>
<fC03 i1="02" i2="X" l="SPA"><s0>Reconocimiento carácter</s0>
<s5>02</s5>
</fC03>
<fC03 i1="03" i2="X" l="FRE"><s0>Reconnaissance automatique</s0>
<s5>03</s5>
</fC03>
<fC03 i1="03" i2="X" l="ENG"><s0>Automatic recognition</s0>
<s5>03</s5>
</fC03>
<fC03 i1="03" i2="X" l="SPA"><s0>Reconocimiento automático</s0>
<s5>03</s5>
</fC03>
<fC03 i1="04" i2="X" l="FRE"><s0>Reconnaissance parole</s0>
<s5>04</s5>
</fC03>
<fC03 i1="04" i2="X" l="ENG"><s0>Speech recognition</s0>
<s5>04</s5>
</fC03>
<fC03 i1="04" i2="X" l="SPA"><s0>Reconocimiento voz</s0>
<s5>04</s5>
</fC03>
<fC03 i1="05" i2="X" l="FRE"><s0>Analyse spectre</s0>
<s5>05</s5>
</fC03>
<fC03 i1="05" i2="X" l="ENG"><s0>Spectrum analysis</s0>
<s5>05</s5>
</fC03>
<fC03 i1="05" i2="X" l="SPA"><s0>Análisis espectro</s0>
<s5>05</s5>
</fC03>
<fC03 i1="06" i2="3" l="FRE"><s0>Analyse cepstrale</s0>
<s5>06</s5>
</fC03>
<fC03 i1="06" i2="3" l="ENG"><s0>Cepstral analysis</s0>
<s5>06</s5>
</fC03>
<fC03 i1="07" i2="X" l="FRE"><s0>Base donnée</s0>
<s5>07</s5>
</fC03>
<fC03 i1="07" i2="X" l="ENG"><s0>Database</s0>
<s5>07</s5>
</fC03>
<fC03 i1="07" i2="X" l="SPA"><s0>Base dato</s0>
<s5>07</s5>
</fC03>
<fC03 i1="08" i2="3" l="FRE"><s0>Modèle Markov variable cachée</s0>
<s5>08</s5>
</fC03>
<fC03 i1="08" i2="3" l="ENG"><s0>Hidden Markov models</s0>
<s5>08</s5>
</fC03>
<fC03 i1="09" i2="X" l="FRE"><s0>Précision</s0>
<s5>09</s5>
</fC03>
<fC03 i1="09" i2="X" l="ENG"><s0>Accuracy</s0>
<s5>09</s5>
</fC03>
<fC03 i1="09" i2="X" l="SPA"><s0>Precisión</s0>
<s5>09</s5>
</fC03>
<fC03 i1="10" i2="X" l="FRE"><s0>Intervalle confiance</s0>
<s5>10</s5>
</fC03>
<fC03 i1="10" i2="X" l="ENG"><s0>Confidence interval</s0>
<s5>10</s5>
</fC03>
<fC03 i1="10" i2="X" l="SPA"><s0>Intervalo confianza</s0>
<s5>10</s5>
</fC03>
<fC03 i1="11" i2="X" l="FRE"><s0>Segmentation</s0>
<s5>11</s5>
</fC03>
<fC03 i1="11" i2="X" l="ENG"><s0>Segmentation</s0>
<s5>11</s5>
</fC03>
<fC03 i1="11" i2="X" l="SPA"><s0>Segmentación</s0>
<s5>11</s5>
</fC03>
<fC03 i1="12" i2="X" l="FRE"><s0>Algorithme</s0>
<s5>12</s5>
</fC03>
<fC03 i1="12" i2="X" l="ENG"><s0>Algorithm</s0>
<s5>12</s5>
</fC03>
<fC03 i1="12" i2="X" l="SPA"><s0>Algoritmo</s0>
<s5>12</s5>
</fC03>
<fC03 i1="13" i2="X" l="FRE"><s0>Implémentation</s0>
<s5>13</s5>
</fC03>
<fC03 i1="13" i2="X" l="ENG"><s0>Implementation</s0>
<s5>13</s5>
</fC03>
<fC03 i1="13" i2="X" l="SPA"><s0>Implementación</s0>
<s5>13</s5>
</fC03>
<fC03 i1="14" i2="X" l="FRE"><s0>Reconnaissance forme</s0>
<s5>14</s5>
</fC03>
<fC03 i1="14" i2="X" l="ENG"><s0>Pattern recognition</s0>
<s5>14</s5>
</fC03>
<fC03 i1="14" i2="X" l="SPA"><s0>Reconocimiento patrón</s0>
<s5>14</s5>
</fC03>
<fC03 i1="15" i2="X" l="FRE"><s0>Traitement parole</s0>
<s5>15</s5>
</fC03>
<fC03 i1="15" i2="X" l="ENG"><s0>Speech processing</s0>
<s5>15</s5>
</fC03>
<fC03 i1="15" i2="X" l="SPA"><s0>Tratamiento palabra</s0>
<s5>15</s5>
</fC03>
<fC03 i1="16" i2="X" l="FRE"><s0>Analyse signal</s0>
<s5>16</s5>
</fC03>
<fC03 i1="16" i2="X" l="ENG"><s0>Signal analysis</s0>
<s5>16</s5>
</fC03>
<fC03 i1="16" i2="X" l="SPA"><s0>Análisis de señal</s0>
<s5>16</s5>
</fC03>
<fC03 i1="17" i2="X" l="FRE"><s0>Approche probabiliste</s0>
<s5>17</s5>
</fC03>
<fC03 i1="17" i2="X" l="ENG"><s0>Probabilistic approach</s0>
<s5>17</s5>
</fC03>
<fC03 i1="17" i2="X" l="SPA"><s0>Enfoque probabilista</s0>
<s5>17</s5>
</fC03>
<fC03 i1="18" i2="3" l="FRE"><s0>Extraction caractéristique</s0>
<s5>18</s5>
</fC03>
<fC03 i1="18" i2="3" l="ENG"><s0>Feature extraction</s0>
<s5>18</s5>
</fC03>
<fC03 i1="19" i2="X" l="FRE"><s0>Traitement signal</s0>
<s5>19</s5>
</fC03>
<fC03 i1="19" i2="X" l="ENG"><s0>Signal processing</s0>
<s5>19</s5>
</fC03>
<fC03 i1="19" i2="X" l="SPA"><s0>Procesamiento señal</s0>
<s5>19</s5>
</fC03>
<fN21><s1>285</s1>
</fN21>
<fN44 i1="01"><s1>OTO</s1>
</fN44>
<fN82><s1>OTO</s1>
</fN82>
</pA>
<pR><fA30 i1="01" i2="1" l="ENG"><s1>International conference on acoustics, speech, and signal processing</s1>
<s3>Orlando FL USA</s3>
<s4>2002-05-13</s4>
</fA30>
</pR>
</standard>
<server><NO>PASCAL 04-0510947 INIST</NO>
<ET>On the use of high order derivatives for high performance alphabet recognition</ET>
<AU>DI MARTINO (Joseph)</AU>
<AF>LORIA, B.P 239/Vandœuvre-lès-Nancy 54506/France (1 aut.)</AF>
<DT>Publication en série; Congrès; Niveau analytique</DT>
<SO>Proceedings of the ... IEEE International Conference on Acoustics, Speech and Signal Processing; ISSN 1520-6149; Etats-Unis; Da. 2002; vol I, 953-956; Bibl. 10 ref.</SO>
<LA>Anglais</LA>
<EA>In this paper I propose new feature vectors for automatic speech recognition. They are based on Mel-cepstrum vectors augmented by derivatives. In the literature, many systems using just two derivatives-delta and delta delta- are described. But none explores the use of higher order derivatives. This paper presents alphabet recognition results on the Isolet database, using feature vectors containing up to the fifth-order derivatives. For this paper I did not use the HTK toolkit proposed by Cambridge University. I developed my own HMM system. I show that with vectors incorporating all the derivatives up to the fifth one, 97.54% mean recognition accuracy was achieved, result which is comparable to the best published one on this database (97.6%), if the recognition accuracy confidence interval concerning this task (approximately 0.3%) is taken into account. It is important to note that this result was obtained without segmenting the speech files by an endpoint detection algorithm. This is an unfavourable experimental condition compared to previous published research works. As a consequence, my system is one of the most powerful systems ever implemented for alphabet recognition.</EA>
<CC>001D04A05A; 001D04A05B; 001D04A04A1</CC>
<FD>Haute performance; Reconnaissance caractère; Reconnaissance automatique; Reconnaissance parole; Analyse spectre; Analyse cepstrale; Base donnée; Modèle Markov variable cachée; Précision; Intervalle confiance; Segmentation; Algorithme; Implémentation; Reconnaissance forme; Traitement parole; Analyse signal; Approche probabiliste; Extraction caractéristique; Traitement signal</FD>
<ED>High performance; Character recognition; Automatic recognition; Speech recognition; Spectrum analysis; Cepstral analysis; Database; Hidden Markov models; Accuracy; Confidence interval; Segmentation; Algorithm; Implementation; Pattern recognition; Speech processing; Signal analysis; Probabilistic approach; Feature extraction; Signal processing</ED>
<SD>Alto rendimiento; Reconocimiento carácter; Reconocimiento automático; Reconocimiento voz; Análisis espectro; Base dato; Precisión; Intervalo confianza; Segmentación; Algoritmo; Implementación; Reconocimiento patrón; Tratamiento palabra; Análisis de señal; Enfoque probabilista; Procesamiento señal</SD>
<LO>INIST-Y 38009.354000117914991195</LO>
<ID>04-0510947</ID>
</server>
</inist>
</record>

Pour manipuler ce document sous Unix (Dilib)

EXPLOR_STEP=$WICRI_ROOT/Wicri/Lorraine/explor/InforLorV4/Data/PascalFrancis/Corpus

HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000627 | SxmlIndent | more

HfdSelect -h $EXPLOR_AREA/Data/PascalFrancis/Corpus/biblio.hfd -nk 000627 | SxmlIndent | more

Pour mettre un lien sur cette page dans le réseau Wicri

{{Explor lien
   |wiki=    Wicri/Lorraine
   |area=    InforLorV4
   |flux=    PascalFrancis
   |étape=   Corpus
   |type=    RBID
   |clé=     Pascal:04-0510947
   |texte=   On the use of high order derivatives for high performance alphabet recognition
}}

This area was generated with Dilib version V0.6.33.
Data generation: Mon Jun 10 21:56:28 2019. Site generation: Fri Feb 25 15:29:27 2022

	Serveur d'exploration sur la recherche en informatique en Lorraine
	Attention, ce site est en cours de développement ! Attention, site généré par des moyens informatiques à partir de corpus bruts. Les informations ne sont donc pas validées.

Serveur d'exploration sur la recherche en informatique en Lorraine

On the use of high order derivatives for high performance alphabet recognition

On the use of high order derivatives for high performance alphabet recognition

Source :

Descripteurs français

English descriptors

Abstract

Notice en format standard (ISO 2709)

Format Inist (serveur)

Links to Exploration step

Le document en format XML

Pour manipuler ce document sous Unix (Dilib)

Pour mettre un lien sur cette page dans le réseau Wicri