Optical character recognition by a neural network
Identifieur interne : 000070 ( Istex/Corpus ); précédent : 000069; suivant : 000071Optical character recognition by a neural network
Auteurs : Michael Sabourin ; Amar MiticheSource :
- Neural Networks [ 0893-6080 ] ; 1992.
English descriptors
Abstract
An optical character recognition (OCR) system, which uses a multilayer perceptron (MLP) neural network classifier, is described. The neural network classifier has the advantage of being fast (highly parallel), easily trainable, and capable of creating arbitrary partitions of the input feature space. Issues in design of the neural network that we examine include the selection of input features, the choice of network learning and momentum parameters, and the selection of training patterns. We also provide a detailed analysis of the learning parameters to provide insight into the MLP, and to suggest a mechanism to automatically tune these parameters. An OCR neural network classifier was trained to recognize characters from a large number of fonts, thereby approaching an omnifont environment. Samples were selected from over 200 fonts and 50 typical office documents, for a total of 110,000 training patterns. In order to evaluate the performance of the MLP classifier, a comparison is made with a high performance dynamic contour warping (DCW) classifier. The base recognition rate on the test set is 96.7% for the neural network classifier, compared to 95.9% for the DCW classifier.
Url:
DOI: 10.1016/S0893-6080(05)80144-3
Links to Exploration step
ISTEX:A31C67A1D4C0AE57B33081612CC57515B54B078ELe document en format XML
<record><TEI wicri:istexFullTextTei="biblStruct"><teiHeader><fileDesc><titleStmt><title xml:lang="en">Optical character recognition by a neural network</title><author><name sortKey="Sabourin, Michael" sort="Sabourin, Michael" uniqKey="Sabourin M" first="Michael" last="Sabourin">Michael Sabourin</name><affiliation><mods:affiliation>INRS-Télécommunications, Canada</mods:affiliation></affiliation></author><author><name sortKey="Mitiche, Amar" sort="Mitiche, Amar" uniqKey="Mitiche A" first="Amar" last="Mitiche">Amar Mitiche</name><affiliation><mods:affiliation>INRS-Télécommunications, Canada</mods:affiliation></affiliation></author></titleStmt><publicationStmt><idno type="wicri:source">ISTEX</idno><idno type="RBID">ISTEX:A31C67A1D4C0AE57B33081612CC57515B54B078E</idno><date when="1992" year="1992">1992</date><idno type="doi">10.1016/S0893-6080(05)80144-3</idno><idno type="url">https://api.istex.fr/document/A31C67A1D4C0AE57B33081612CC57515B54B078E/fulltext/pdf</idno><idno type="wicri:Area/Istex/Corpus">000070</idno></publicationStmt><sourceDesc><biblStruct><analytic><title level="a" type="main" xml:lang="en">Optical character recognition by a neural network</title><author><name sortKey="Sabourin, Michael" sort="Sabourin, Michael" uniqKey="Sabourin M" first="Michael" last="Sabourin">Michael Sabourin</name><affiliation><mods:affiliation>INRS-Télécommunications, Canada</mods:affiliation></affiliation></author><author><name sortKey="Mitiche, Amar" sort="Mitiche, Amar" uniqKey="Mitiche A" first="Amar" last="Mitiche">Amar Mitiche</name><affiliation><mods:affiliation>INRS-Télécommunications, Canada</mods:affiliation></affiliation></author></analytic><monogr></monogr><series><title level="j">Neural Networks</title><title level="j" type="abbrev">NN</title><idno type="ISSN">0893-6080</idno><imprint><publisher>ELSEVIER</publisher><date type="published" when="1992">1992</date><biblScope unit="volume">5</biblScope><biblScope unit="issue">5</biblScope><biblScope unit="page" from="843">843</biblScope><biblScope unit="page" to="852">852</biblScope></imprint><idno type="ISSN">0893-6080</idno></series><idno type="istex">A31C67A1D4C0AE57B33081612CC57515B54B078E</idno><idno type="DOI">10.1016/S0893-6080(05)80144-3</idno><idno type="PII">S0893-6080(05)80144-3</idno><idno type="ArticleID">05801443</idno></biblStruct></sourceDesc><seriesStmt><idno type="ISSN">0893-6080</idno></seriesStmt></fileDesc><profileDesc><textClass><keywords scheme="KwdEn" xml:lang="en"><term>Hierarchical classifier</term><term>Learning parameters</term><term>Momentum</term><term>Multilayer perceptron</term><term>OCR</term></keywords></textClass><langUsage><language ident="en">en</language></langUsage></profileDesc></teiHeader><front><div type="abstract" xml:lang="en">An optical character recognition (OCR) system, which uses a multilayer perceptron (MLP) neural network classifier, is described. The neural network classifier has the advantage of being fast (highly parallel), easily trainable, and capable of creating arbitrary partitions of the input feature space. Issues in design of the neural network that we examine include the selection of input features, the choice of network learning and momentum parameters, and the selection of training patterns. We also provide a detailed analysis of the learning parameters to provide insight into the MLP, and to suggest a mechanism to automatically tune these parameters. An OCR neural network classifier was trained to recognize characters from a large number of fonts, thereby approaching an omnifont environment. Samples were selected from over 200 fonts and 50 typical office documents, for a total of 110,000 training patterns. In order to evaluate the performance of the MLP classifier, a comparison is made with a high performance dynamic contour warping (DCW) classifier. The base recognition rate on the test set is 96.7% for the neural network classifier, compared to 95.9% for the DCW classifier.</div></front></TEI><istex><corpusName>elsevier</corpusName><author><json:item><name>Michael Sabourin</name><affiliations><json:string>INRS-Télécommunications, Canada</json:string></affiliations></json:item><json:item><name>Amar Mitiche</name><affiliations><json:string>INRS-Télécommunications, Canada</json:string></affiliations></json:item></author><subject><json:item><lang><json:string>eng</json:string></lang><value>OCR</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Multilayer perceptron</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Learning parameters</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Momentum</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Hierarchical classifier</value></json:item></subject><articleId><json:string>05801443</json:string></articleId><language><json:string>eng</json:string></language><abstract>An optical character recognition (OCR) system, which uses a multilayer perceptron (MLP) neural network classifier, is described. The neural network classifier has the advantage of being fast (highly parallel), easily trainable, and capable of creating arbitrary partitions of the input feature space. Issues in design of the neural network that we examine include the selection of input features, the choice of network learning and momentum parameters, and the selection of training patterns. We also provide a detailed analysis of the learning parameters to provide insight into the MLP, and to suggest a mechanism to automatically tune these parameters. An OCR neural network classifier was trained to recognize characters from a large number of fonts, thereby approaching an omnifont environment. Samples were selected from over 200 fonts and 50 typical office documents, for a total of 110,000 training patterns. In order to evaluate the performance of the MLP classifier, a comparison is made with a high performance dynamic contour warping (DCW) classifier. The base recognition rate on the test set is 96.7% for the neural network classifier, compared to 95.9% for the DCW classifier.</abstract><qualityIndicators><score>7.196</score><pdfVersion>1.3</pdfVersion><pdfPageSize>612 x 792 pts (letter)</pdfPageSize><refBibsNative>true</refBibsNative><keywordCount>5</keywordCount><abstractCharCount>1191</abstractCharCount><pdfWordCount>5362</pdfWordCount><pdfCharCount>32076</pdfCharCount><pdfPageCount>10</pdfPageCount><abstractWordCount>183</abstractWordCount></qualityIndicators><title>Optical character recognition by a neural network</title><pii><json:string>S0893-6080(05)80144-3</json:string></pii><genre><json:string>research-article</json:string></genre><serie><genre></genre><language><json:string>unknown</json:string></language><title>unpubished M.Eng.</title></serie><host><volume>5</volume><pii><json:string>S0893-6080(05)X8134-5</json:string></pii><pages><last>852</last><first>843</first></pages><issn><json:string>0893-6080</json:string></issn><issue>5</issue><genre><json:string>Journal</json:string></genre><language><json:string>unknown</json:string></language><title>Neural Networks</title><publicationDate>1992</publicationDate></host><categories><wos><json:string>COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE</json:string><json:string>NEUROSCIENCES</json:string></wos></categories><publicationDate>1992</publicationDate><copyrightDate>1992</copyrightDate><doi><json:string>10.1016/S0893-6080(05)80144-3</json:string></doi><id>A31C67A1D4C0AE57B33081612CC57515B54B078E</id><fulltext><json:item><original>true</original><mimetype>application/pdf</mimetype><extension>pdf</extension><uri>https://api.istex.fr/document/A31C67A1D4C0AE57B33081612CC57515B54B078E/fulltext/pdf</uri></json:item><json:item><original>true</original><mimetype>text/plain</mimetype><extension>txt</extension><uri>https://api.istex.fr/document/A31C67A1D4C0AE57B33081612CC57515B54B078E/fulltext/txt</uri></json:item><json:item><original>false</original><mimetype>application/zip</mimetype><extension>zip</extension><uri>https://api.istex.fr/document/A31C67A1D4C0AE57B33081612CC57515B54B078E/fulltext/zip</uri></json:item><istex:fulltextTEI uri="https://api.istex.fr/document/A31C67A1D4C0AE57B33081612CC57515B54B078E/fulltext/tei"><teiHeader><fileDesc><titleStmt><title level="a" type="main" xml:lang="en">Optical character recognition by a neural network</title></titleStmt><publicationStmt><authority>ISTEX</authority><publisher>ELSEVIER</publisher><availability><p>ELSEVIER</p></availability><date>1992</date></publicationStmt><notesStmt><note type="content">Section title: Original Contribution</note></notesStmt><sourceDesc><biblStruct type="inbook"><analytic><title level="a" type="main" xml:lang="en">Optical character recognition by a neural network</title><author><persName><forename type="first">Michael</forename><surname>Sabourin</surname></persName><note type="biography">Bell-Northern Research</note><affiliation>Bell-Northern Research</affiliation><affiliation>INRS-Télécommunications, Canada</affiliation></author><author><persName><forename type="first">Amar</forename><surname>Mitiche</surname></persName><note type="biography">Requests for reprints should be sent to Amar Mitiche, INRS-Télécommunications, 3 Place du Commerce, Ile des Soeurs, Verdun, Que., Canada, H3E 1H6.</note><affiliation>Requests for reprints should be sent to Amar Mitiche, INRS-Télécommunications, 3 Place du Commerce, Ile des Soeurs, Verdun, Que., Canada, H3E 1H6.</affiliation><affiliation>INRS-Télécommunications, Canada</affiliation></author></analytic><monogr><title level="j">Neural Networks</title><title level="j" type="abbrev">NN</title><idno type="pISSN">0893-6080</idno><idno type="PII">S0893-6080(05)X8134-5</idno><imprint><publisher>ELSEVIER</publisher><date type="published" when="1992"></date><biblScope unit="volume">5</biblScope><biblScope unit="issue">5</biblScope><biblScope unit="page" from="843">843</biblScope><biblScope unit="page" to="852">852</biblScope></imprint></monogr><idno type="istex">A31C67A1D4C0AE57B33081612CC57515B54B078E</idno><idno type="DOI">10.1016/S0893-6080(05)80144-3</idno><idno type="PII">S0893-6080(05)80144-3</idno><idno type="ArticleID">05801443</idno></biblStruct></sourceDesc></fileDesc><profileDesc><creation><date>1992</date></creation><langUsage><language ident="en">en</language></langUsage><abstract xml:lang="en"><p>An optical character recognition (OCR) system, which uses a multilayer perceptron (MLP) neural network classifier, is described. The neural network classifier has the advantage of being fast (highly parallel), easily trainable, and capable of creating arbitrary partitions of the input feature space. Issues in design of the neural network that we examine include the selection of input features, the choice of network learning and momentum parameters, and the selection of training patterns. We also provide a detailed analysis of the learning parameters to provide insight into the MLP, and to suggest a mechanism to automatically tune these parameters. An OCR neural network classifier was trained to recognize characters from a large number of fonts, thereby approaching an omnifont environment. Samples were selected from over 200 fonts and 50 typical office documents, for a total of 110,000 training patterns. In order to evaluate the performance of the MLP classifier, a comparison is made with a high performance dynamic contour warping (DCW) classifier. The base recognition rate on the test set is 96.7% for the neural network classifier, compared to 95.9% for the DCW classifier.</p></abstract><textClass xml:lang="en"><keywords scheme="keyword"><list><head>Keywords</head><item><term>OCR</term></item><item><term>Multilayer perceptron</term></item><item><term>Learning parameters</term></item><item><term>Momentum</term></item><item><term>Hierarchical classifier</term></item></list></keywords></textClass></profileDesc><revisionDesc><change when="1992-01-16">Registration</change><change when="1992">Published</change></revisionDesc></teiHeader></istex:fulltextTEI></fulltext><metadata><istex:metadataXml wicri:clean="Elsevier doc found" wicri:toSee="Elsevier, no converted or simple article"><istex:xmlDeclaration>version="1.0" encoding="utf-8"</istex:xmlDeclaration><istex:docType PUBLIC="-//ES//DTD journal article DTD version 5.0.1//EN//XML" URI="art501.dtd" name="istex:docType"></istex:docType><istex:document><article version="5.0" xml:lang="en" docsubtype="fla"><item-info><jid>NN</jid><aid>05801443</aid><ce:pii>S0893-6080(05)80144-3</ce:pii><ce:doi>10.1016/S0893-6080(05)80144-3</ce:doi><ce:copyright type="unknown" year="1992">Pergamon Press Ltd.</ce:copyright></item-info><head><ce:dochead><ce:textfn>Original Contribution</ce:textfn></ce:dochead><ce:title>Optical character recognition by a neural network</ce:title><ce:author-group><ce:author><ce:given-name>Michael</ce:given-name><ce:surname>Sabourin</ce:surname><ce:cross-ref refid="fn1"><ce:sup>1</ce:sup></ce:cross-ref></ce:author><ce:author><ce:given-name>Amar</ce:given-name><ce:surname>Mitiche</ce:surname><ce:cross-ref refid="cor1"><ce:sup>2</ce:sup></ce:cross-ref></ce:author><ce:affiliation id="aff1"><ce:textfn>INRS-Télécommunications, Canada</ce:textfn></ce:affiliation><ce:correspondence id="cor1"><ce:label>2</ce:label><ce:text>Requests for reprints should be sent to Amar Mitiche, INRS-Télécommunications, 3 Place du Commerce, Ile des Soeurs, Verdun, Que., Canada, H3E 1H6.</ce:text></ce:correspondence><ce:footnote id="fn1"><ce:label>1</ce:label><ce:note-para>Bell-Northern Research</ce:note-para></ce:footnote></ce:author-group><ce:date-received day="8" month="3" year="1991"></ce:date-received><ce:date-accepted day="16" month="1" year="1992"></ce:date-accepted><ce:abstract id="ab1" class="author" xml:lang="en"><ce:section-title>Abstract</ce:section-title><ce:abstract-sec><ce:simple-para>An optical character recognition (OCR) system, which uses a multilayer perceptron (MLP) neural network classifier, is described. The neural network classifier has the advantage of being fast (highly parallel), easily trainable, and capable of creating arbitrary partitions of the input feature space. Issues in design of the neural network that we examine include the selection of input features, the choice of network learning and momentum parameters, and the selection of training patterns. We also provide a detailed analysis of the learning parameters to provide insight into the MLP, and to suggest a mechanism to automatically tune these parameters. An OCR neural network classifier was trained to recognize characters from a large number of fonts, thereby approaching an omnifont environment. Samples were selected from over 200 fonts and 50 typical office documents, for a total of 110,000 training patterns. In order to evaluate the performance of the MLP classifier, a comparison is made with a high performance dynamic contour warping (DCW) classifier. The base recognition rate on the test set is 96.7% for the neural network classifier, compared to 95.9% for the DCW classifier.</ce:simple-para></ce:abstract-sec></ce:abstract><ce:keywords class="keyword" xml:lang="en"><ce:section-title>Keywords</ce:section-title><ce:keyword><ce:text>OCR</ce:text></ce:keyword><ce:keyword><ce:text>Multilayer perceptron</ce:text></ce:keyword><ce:keyword><ce:text>Learning parameters</ce:text></ce:keyword><ce:keyword><ce:text>Momentum</ce:text></ce:keyword><ce:keyword><ce:text>Hierarchical classifier</ce:text></ce:keyword></ce:keywords></head><tail><ce:bibliography><ce:section-title>References</ce:section-title><ce:bibliography-sec><ce:bib-reference id="bib1"><ce:label>Baird and Thompson, 1987</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Baird</ce:surname><ce:given-name>H.S.</ce:given-name></sb:author><sb:author><ce:surname>Thompson</ce:surname><ce:given-name>K.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Reading chess</sb:maintitle></sb:title></sb:contribution><sb:host><sb:edited-book><sb:title><sb:maintitle>Proceedings of the IEEE Computer Society Workshop on Computer Vision</sb:maintitle></sb:title><sb:date>1987</sb:date></sb:edited-book><sb:pages><sb:first-page>277</sb:first-page><sb:last-page>282</sb:last-page></sb:pages></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib2"><ce:label>Cash and Hatamian, 1987</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Cash</ce:surname><ce:given-name>G.L.</ce:given-name></sb:author><sb:author><ce:surname>Hatamian</ce:surname><ce:given-name>M.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Optical character recognition by the method of moments</sb:maintitle></sb:title></sb:contribution><sb:host><sb:issue><sb:series><sb:title><sb:maintitle>Computer Vision, Graphics and Image Processing</sb:maintitle></sb:title><sb:volume-nr>39</sb:volume-nr></sb:series><sb:date>1987</sb:date></sb:issue><sb:pages><sb:first-page>291</sb:first-page><sb:last-page>310</sb:last-page></sb:pages></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib3"><ce:label>Duffie, 1985</ce:label><sb:reference><sb:comment>unpubished M.Eng.</sb:comment><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Duffie</ce:surname><ce:given-name>P.K.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Contour elastic matching for omni-font character recognition</sb:maintitle></sb:title></sb:contribution><sb:host><sb:edited-book><sb:book-series><sb:series><sb:title><sb:maintitle>Thesis</sb:maintitle></sb:title></sb:series></sb:book-series><sb:date>1985</sb:date><sb:publisher><sb:name>McGill University</sb:name><sb:location>New York</sb:location></sb:publisher></sb:edited-book></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib4"><ce:label>Haykin, 1978</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Haykin</ce:surname><ce:given-name>S.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Communication systems</sb:maintitle></sb:title></sb:contribution><sb:host><sb:book><sb:date>1978</sb:date><sb:publisher><sb:name>John Wiley & Sons</sb:name><sb:location>New York</sb:location></sb:publisher></sb:book></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib5"><ce:label>Kahan, et al, 1987</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Kahan</ce:surname><ce:given-name>S.</ce:given-name></sb:author><sb:author><ce:surname>Pavlidis</ce:surname><ce:given-name>T.</ce:given-name></sb:author><sb:author><ce:surname>Baird</ce:surname><ce:given-name>H.S.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>On the recognition of printed characters of any font and size</sb:maintitle></sb:title></sb:contribution><sb:host><sb:issue><sb:series><sb:title><sb:maintitle>IEEE Pattern Analysis and Machine Intelligence</sb:maintitle></sb:title><sb:volume-nr>9</sb:volume-nr></sb:series><sb:date>1987</sb:date></sb:issue><sb:pages><sb:first-page>274</sb:first-page><sb:last-page>287</sb:last-page></sb:pages></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib6"><ce:label>Kohonen, 1988</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Kohonen</ce:surname><ce:given-name>T.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Self-organization and associative memory</sb:maintitle></sb:title></sb:contribution><sb:host><sb:edited-book><sb:edition>2nd ed.</sb:edition><sb:date>1988</sb:date><sb:publisher><sb:name>Springer-Verlag</sb:name><sb:location>Berlin</sb:location></sb:publisher></sb:edited-book></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib7"><ce:label>Lippman, 1987</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Lippman</ce:surname><ce:given-name>R.P.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>An introduction to computing with neural networks</sb:maintitle></sb:title></sb:contribution><sb:host><sb:edited-book><sb:book-series><sb:series><sb:title><sb:maintitle>IEEE Acoustics, Speech, and Signal Processing Magazine</sb:maintitle></sb:title></sb:series></sb:book-series><sb:date>1987</sb:date></sb:edited-book><sb:pages><sb:first-page>4</sb:first-page><sb:last-page>22</sb:last-page></sb:pages></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib8"><ce:label>Minsky and Papert, 1988</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Minsky</ce:surname><ce:given-name>M.L.</ce:given-name></sb:author><sb:author><ce:surname>Papert</ce:surname><ce:given-name>S.A.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Perceptrons. An introduction to computational geometry</sb:maintitle></sb:title></sb:contribution><sb:host><sb:book><sb:date>1988</sb:date><sb:publisher><sb:name>MIT Press</sb:name><sb:location>Cambridge, MA</sb:location></sb:publisher></sb:book></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib9"><ce:label>Roberts and Agarwal, 1990</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Roberts</ce:surname><ce:given-name>G.</ce:given-name></sb:author><sb:author><ce:surname>Agarwal</ce:surname><ce:given-name>V.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Course notes: VLSI methods for neural networks</sb:maintitle></sb:title></sb:contribution><sb:host><sb:book><sb:date>1990</sb:date><sb:publisher><sb:name>Department Electrical Engineering, McGill University</sb:name></sb:publisher></sb:book></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib10"><ce:label>Robillard, 1990</ce:label><sb:reference><sb:comment>unpublished Masters</sb:comment><sb:contribution langtype="iso" xml:lang="fr"><sb:authors><sb:author><ce:surname>Robillard</ce:surname><ce:given-name>S.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Contribution à l'étude des mémoires associatives</sb:maintitle></sb:title></sb:contribution><sb:host><sb:edited-book><sb:book-series><sb:series><sb:title><sb:maintitle>Thesis</sb:maintitle></sb:title></sb:series></sb:book-series><sb:date>1990</sb:date><sb:publisher><sb:name>INRS-Telecommunications</sb:name><sb:location>Montreal, Canada</sb:location></sb:publisher></sb:edited-book></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib11"><ce:label>Rumelhart and McClelland, 1986</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Rumelhart</ce:surname><ce:given-name>D.E.</ce:given-name></sb:author><sb:author><ce:surname>McClelland</ce:surname><ce:given-name>J.L.</ce:given-name></sb:author></sb:authors></sb:contribution><sb:host><sb:edited-book><sb:book-series><sb:series><sb:title><sb:maintitle>Parallel distributed processing</sb:maintitle></sb:title><sb:volume-nr>vol I and II</sb:volume-nr></sb:series></sb:book-series><sb:date>1986</sb:date><sb:publisher><sb:name>MIT Press</sb:name><sb:location>Cambridge, MA</sb:location></sb:publisher></sb:edited-book></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib12"><ce:label>Sinha and Prasada, 1987</ce:label><sb:reference><sb:comment>May</sb:comment><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Sinha</ce:surname><ce:given-name>R.M.K.</ce:given-name></sb:author><sb:author><ce:surname>Prasada</ce:surname><ce:given-name>B.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>An integrated omni-font text recognition system</sb:maintitle></sb:title></sb:contribution><sb:host><sb:issue><sb:series><sb:title><sb:maintitle>Rapport Tech. INRS-Telecommunications, No. 87-21</sb:maintitle></sb:title></sb:series><sb:date>1987</sb:date></sb:issue></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib13"><ce:label>Sinha and Prasada, 1988</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Sinha</ce:surname><ce:given-name>R.M.K.</ce:given-name></sb:author><sb:author><ce:surname>Prasada</ce:surname><ce:given-name>B.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Visual text recognition through contextual processing</sb:maintitle></sb:title></sb:contribution><sb:host><sb:issue><sb:series><sb:title><sb:maintitle>Pattern Recognition</sb:maintitle></sb:title><sb:volume-nr>21</sb:volume-nr></sb:series><sb:date>1988</sb:date></sb:issue><sb:pages><sb:first-page>463</sb:first-page><sb:last-page>479</sb:last-page></sb:pages></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib14"><ce:label>Sinha,</ce:label><ce:other-ref><ce:textref>Sinha, R. M. K., Prasada, B., Houle, G., & Sabourin, M. (in press). Hybrid contextual text recognition with string matching. <ce:italic>IEEE Pattern Analysis and Machine Intelligence</ce:italic>.</ce:textref></ce:other-ref></ce:bib-reference><ce:bib-reference id="bib15"><ce:label>Suen, et al, 1980</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Suen</ce:surname><ce:given-name>C.Y.</ce:given-name></sb:author><sb:author><ce:surname>Berthod</ce:surname><ce:given-name>M.</ce:given-name></sb:author><sb:author><ce:surname>Mori</ce:surname><ce:given-name>S.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Automatic recognition of handprinted characters—The state of the art</sb:maintitle></sb:title></sb:contribution><sb:host><sb:edited-book><sb:book-series><sb:series><sb:title><sb:maintitle>Proceedings of the IEEE</sb:maintitle></sb:title><sb:volume-nr>68</sb:volume-nr></sb:series></sb:book-series><sb:date>1980</sb:date></sb:edited-book><sb:pages><sb:first-page>469</sb:first-page><sb:last-page>487</sb:last-page></sb:pages></sb:host></sb:reference></ce:bib-reference><ce:bib-reference id="bib16"><ce:label>Tappert, 1982</ce:label><sb:reference><sb:contribution langtype="en"><sb:authors><sb:author><ce:surname>Tappert</ce:surname><ce:given-name>C.C.</ce:given-name></sb:author></sb:authors><sb:title><sb:maintitle>Cursive script recognition by elastic matching</sb:maintitle></sb:title></sb:contribution><sb:host><sb:issue><sb:series><sb:title><sb:maintitle>IBM Journal of Research</sb:maintitle></sb:title><sb:volume-nr>10</sb:volume-nr></sb:series><sb:date>1982</sb:date></sb:issue><sb:pages><sb:first-page>765</sb:first-page><sb:last-page>771</sb:last-page></sb:pages></sb:host></sb:reference></ce:bib-reference></ce:bibliography-sec></ce:bibliography></tail></article></istex:document></istex:metadataXml><mods version="3.6"><titleInfo lang="en"><title>Optical character recognition by a neural network</title></titleInfo><titleInfo type="alternative" lang="en" contentType="CDATA"><title>Optical character recognition by a neural network</title></titleInfo><name type="personal"><namePart type="given">Michael</namePart><namePart type="family">Sabourin</namePart><affiliation>INRS-Télécommunications, Canada</affiliation><description>Bell-Northern Research</description><role><roleTerm type="text">author</roleTerm></role></name><name type="personal"><namePart type="given">Amar</namePart><namePart type="family">Mitiche</namePart><affiliation>INRS-Télécommunications, Canada</affiliation><description>Requests for reprints should be sent to Amar Mitiche, INRS-Télécommunications, 3 Place du Commerce, Ile des Soeurs, Verdun, Que., Canada, H3E 1H6.</description><role><roleTerm type="text">author</roleTerm></role></name><typeOfResource>text</typeOfResource><genre type="research-article" displayLabel="Full-length article"></genre><originInfo><publisher>ELSEVIER</publisher><dateIssued encoding="w3cdtf">1992</dateIssued><dateValid encoding="w3cdtf">1992-01-16</dateValid><copyrightDate encoding="w3cdtf">1992</copyrightDate></originInfo><language><languageTerm type="code" authority="iso639-2b">eng</languageTerm><languageTerm type="code" authority="rfc3066">en</languageTerm></language><physicalDescription><internetMediaType>text/html</internetMediaType></physicalDescription><abstract lang="en">An optical character recognition (OCR) system, which uses a multilayer perceptron (MLP) neural network classifier, is described. The neural network classifier has the advantage of being fast (highly parallel), easily trainable, and capable of creating arbitrary partitions of the input feature space. Issues in design of the neural network that we examine include the selection of input features, the choice of network learning and momentum parameters, and the selection of training patterns. We also provide a detailed analysis of the learning parameters to provide insight into the MLP, and to suggest a mechanism to automatically tune these parameters. An OCR neural network classifier was trained to recognize characters from a large number of fonts, thereby approaching an omnifont environment. Samples were selected from over 200 fonts and 50 typical office documents, for a total of 110,000 training patterns. In order to evaluate the performance of the MLP classifier, a comparison is made with a high performance dynamic contour warping (DCW) classifier. The base recognition rate on the test set is 96.7% for the neural network classifier, compared to 95.9% for the DCW classifier.</abstract><note type="content">Section title: Original Contribution</note><subject lang="en"><genre>Keywords</genre><topic>OCR</topic><topic>Multilayer perceptron</topic><topic>Learning parameters</topic><topic>Momentum</topic><topic>Hierarchical classifier</topic></subject><relatedItem type="host"><titleInfo><title>Neural Networks</title></titleInfo><titleInfo type="abbreviated"><title>NN</title></titleInfo><genre type="Journal">journal</genre><originInfo><dateIssued encoding="w3cdtf">199209</dateIssued></originInfo><identifier type="ISSN">0893-6080</identifier><identifier type="PII">S0893-6080(05)X8134-5</identifier><part><detail type="volume"><number>5</number><caption>vol.</caption></detail><detail type="issue"><number>5</number><caption>no.</caption></detail><extent unit="issue pages"><start>735</start><end>867</end></extent><extent unit="pages"><start>843</start><end>852</end></extent></part></relatedItem><identifier type="istex">A31C67A1D4C0AE57B33081612CC57515B54B078E</identifier><identifier type="DOI">10.1016/S0893-6080(05)80144-3</identifier><identifier type="PII">S0893-6080(05)80144-3</identifier><identifier type="ArticleID">05801443</identifier><accessCondition type="use and reproduction" contentType="">© 1992Pergamon Press Ltd.</accessCondition><recordInfo><recordContentSource>ELSEVIER</recordContentSource><recordOrigin>Pergamon Press Ltd., ©1992</recordOrigin></recordInfo></mods></metadata><enrichments><istex:catWosTEI uri="https://api.istex.fr/document/A31C67A1D4C0AE57B33081612CC57515B54B078E/enrichments/catWos"><teiHeader><profileDesc><textClass><classCode scheme="WOS">COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE</classCode><classCode scheme="WOS">NEUROSCIENCES</classCode></textClass></profileDesc></teiHeader></istex:catWosTEI></enrichments></istex></record>Pour manipuler ce document sous Unix (Dilib)
EXPLOR_STEP=$WICRI_ROOT/Ticri/CIDE/explor/OcrV1/Data/Istex/Corpus
HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000070 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Istex/Corpus/biblio.hfd -nk 000070 | SxmlIndent | more
Pour mettre un lien sur cette page dans le réseau Wicri
{{Explor lien
|wiki= Ticri/CIDE
|area= OcrV1
|flux= Istex
|étape= Corpus
|type= RBID
|clé= ISTEX:A31C67A1D4C0AE57B33081612CC57515B54B078E
|texte= Optical character recognition by a neural network
}}
| This area was generated with Dilib version V0.6.32. |  |