Off-line cursive handwriting recognition using hidden markov models
Identifieur interne : 001029 ( Istex/Corpus ); précédent : 001028; suivant : 001030Off-line cursive handwriting recognition using hidden markov models
Auteurs : H. Bunke ; M. Roth ; E. G. Schukat-TalamazziniSource :
- Pattern Recognition [ 0031-3203 ] ; 1994.
Abstract
A method for the off-line recognition of cursive handwriting based on hidden Markov models (HMMs) is described. The features used in the HMMs are based on the arcs of skeleton graphs of the words to be recognized. An algorithm is applied to the skeleton graph of a word that extracts the edges in a particular order. Given the sequence of edges extracted from the skeleton graph, each edge is transformed into a 10-dimensional feature vector. The features represent information about the location of an edge relative to the four reference lines, its curvature and the degree of the nodes incident to the considered edge. The linear model was adopted as basic HMM topology. Each letter of the alphabet is represented by a linear HMM. Given a dictionary of fixed size, an HMM for each dictionary word is built by sequential concatenation of the HMMs representing the individual letters of the word. Training of the HMMs is done by means of the Baum-Welch algorithm, while the Viterbi algorithm is used for recognition. An average correct recognition rate of over 98% on the word level has been achieved in experiments with cooperative writers using two dictionaries of I50 words each.
Url:
DOI: 10.1016/0031-3203(95)00013-P
Links to Exploration step
ISTEX:2C08A6F9965E87A5FA4832C61259974CFEA9F208Le document en format XML
<record><TEI wicri:istexFullTextTei="biblStruct"><teiHeader><fileDesc><titleStmt><title>Off-line cursive handwriting recognition using hidden markov models</title><author><name sortKey="Bunke, H" sort="Bunke, H" uniqKey="Bunke H" first="H." last="Bunke">H. Bunke</name><affiliation><mods:affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</mods:affiliation></affiliation></author><author><name sortKey="Roth, M" sort="Roth, M" uniqKey="Roth M" first="M." last="Roth">M. Roth</name><affiliation><mods:affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</mods:affiliation></affiliation></author><author><name sortKey="Schukat Talamazzini, E G" sort="Schukat Talamazzini, E G" uniqKey="Schukat Talamazzini E" first="E. G." last="Schukat-Talamazzini">E. G. Schukat-Talamazzini</name><affiliation><mods:affiliation>Lehrstuhl für Informatik 5 (Mustererkennung), Friedrich-Alexander-Universität Erlangen-Nürnberg, Martensstrasse 3, D-91056 Erlangen, Germany</mods:affiliation></affiliation></author></titleStmt><publicationStmt><idno type="wicri:source">ISTEX</idno><idno type="RBID">ISTEX:2C08A6F9965E87A5FA4832C61259974CFEA9F208</idno><date when="1995" year="1995">1995</date><idno type="doi">10.1016/0031-3203(95)00013-P</idno><idno type="url">https://api.istex.fr/document/2C08A6F9965E87A5FA4832C61259974CFEA9F208/fulltext/pdf</idno><idno type="wicri:Area/Istex/Corpus">001029</idno></publicationStmt><sourceDesc><biblStruct><analytic><title level="a">Off-line cursive handwriting recognition using hidden markov models</title><author><name sortKey="Bunke, H" sort="Bunke, H" uniqKey="Bunke H" first="H." last="Bunke">H. Bunke</name><affiliation><mods:affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</mods:affiliation></affiliation></author><author><name sortKey="Roth, M" sort="Roth, M" uniqKey="Roth M" first="M." last="Roth">M. Roth</name><affiliation><mods:affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</mods:affiliation></affiliation></author><author><name sortKey="Schukat Talamazzini, E G" sort="Schukat Talamazzini, E G" uniqKey="Schukat Talamazzini E" first="E. G." last="Schukat-Talamazzini">E. G. Schukat-Talamazzini</name><affiliation><mods:affiliation>Lehrstuhl für Informatik 5 (Mustererkennung), Friedrich-Alexander-Universität Erlangen-Nürnberg, Martensstrasse 3, D-91056 Erlangen, Germany</mods:affiliation></affiliation></author></analytic><monogr></monogr><series><title level="j">Pattern Recognition</title><title level="j" type="abbrev">PR</title><idno type="ISSN">0031-3203</idno><imprint><publisher>ELSEVIER</publisher><date type="published" when="1994">1994</date><biblScope unit="volume">28</biblScope><biblScope unit="issue">9</biblScope><biblScope unit="page" from="1399">1399</biblScope><biblScope unit="page" to="1413">1413</biblScope></imprint><idno type="ISSN">0031-3203</idno></series><idno type="istex">2C08A6F9965E87A5FA4832C61259974CFEA9F208</idno><idno type="DOI">10.1016/0031-3203(95)00013-P</idno><idno type="PII">0031-3203(95)00013-P</idno></biblStruct></sourceDesc><seriesStmt><idno type="ISSN">0031-3203</idno></seriesStmt></fileDesc><profileDesc><textClass></textClass><langUsage><language ident="en">en</language></langUsage></profileDesc></teiHeader><front><div type="abstract" xml:lang="en">A method for the off-line recognition of cursive handwriting based on hidden Markov models (HMMs) is described. The features used in the HMMs are based on the arcs of skeleton graphs of the words to be recognized. An algorithm is applied to the skeleton graph of a word that extracts the edges in a particular order. Given the sequence of edges extracted from the skeleton graph, each edge is transformed into a 10-dimensional feature vector. The features represent information about the location of an edge relative to the four reference lines, its curvature and the degree of the nodes incident to the considered edge. The linear model was adopted as basic HMM topology. Each letter of the alphabet is represented by a linear HMM. Given a dictionary of fixed size, an HMM for each dictionary word is built by sequential concatenation of the HMMs representing the individual letters of the word. Training of the HMMs is done by means of the Baum-Welch algorithm, while the Viterbi algorithm is used for recognition. An average correct recognition rate of over 98% on the word level has been achieved in experiments with cooperative writers using two dictionaries of I50 words each.</div></front></TEI><istex><corpusName>elsevier</corpusName><author><json:item><name>H. Bunke</name><affiliations><json:string>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</json:string></affiliations></json:item><json:item><name>M. Roth</name><affiliations><json:string>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</json:string></affiliations></json:item><json:item><name>E.G. Schukat-Talamazzini</name><affiliations><json:string>Lehrstuhl für Informatik 5 (Mustererkennung), Friedrich-Alexander-Universität Erlangen-Nürnberg, Martensstrasse 3, D-91056 Erlangen, Germany</json:string></affiliations></json:item></author><subject><json:item><lang><json:string>eng</json:string></lang><value>Optical character recognition</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Cursive script recognition</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Off-line recognition</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Hidden Markov model</value></json:item><json:item><lang><json:string>eng</json:string></lang><value>Skeleton graphs</value></json:item></subject><language><json:string>eng</json:string></language><abstract>A method for the off-line recognition of cursive handwriting based on hidden Markov models (HMMs) is described. The features used in the HMMs are based on the arcs of skeleton graphs of the words to be recognized. An algorithm is applied to the skeleton graph of a word that extracts the edges in a particular order. Given the sequence of edges extracted from the skeleton graph, each edge is transformed into a 10-dimensional feature vector. The features represent information about the location of an edge relative to the four reference lines, its curvature and the degree of the nodes incident to the considered edge. The linear model was adopted as basic HMM topology. Each letter of the alphabet is represented by a linear HMM. Given a dictionary of fixed size, an HMM for each dictionary word is built by sequential concatenation of the HMMs representing the individual letters of the word. Training of the HMMs is done by means of the Baum-Welch algorithm, while the Viterbi algorithm is used for recognition. An average correct recognition rate of over 98% on the word level has been achieved in experiments with cooperative writers using two dictionaries of I50 words each.</abstract><qualityIndicators><score>7.376</score><pdfVersion>1.2</pdfVersion><pdfPageSize>533 x 749 pts</pdfPageSize><refBibsNative>true</refBibsNative><keywordCount>5</keywordCount><abstractCharCount>1182</abstractCharCount><pdfWordCount>10309</pdfWordCount><pdfCharCount>50454</pdfCharCount><pdfPageCount>15</pdfPageCount><abstractWordCount>198</abstractWordCount></qualityIndicators><title>Off-line cursive handwriting recognition using hidden markov models</title><pii><json:string>0031-3203(95)00013-P</json:string></pii><genre><json:string>research-article</json:string></genre><serie><genre></genre><language><json:string>unknown</json:string></language><title>7</title></serie><host><volume>28</volume><pii><json:string>S0031-3203(00)X0011-8</json:string></pii><pages><last>1413</last><first>1399</first></pages><issn><json:string>0031-3203</json:string></issn><issue>9</issue><genre><json:string>Journal</json:string></genre><language><json:string>unknown</json:string></language><title>Pattern Recognition</title><publicationDate>1995</publicationDate></host><categories><wos><json:string>COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE</json:string><json:string>ENGINEERING, ELECTRICAL & ELECTRONIC</json:string></wos></categories><publicationDate>1994</publicationDate><copyrightDate>1995</copyrightDate><doi><json:string>10.1016/0031-3203(95)00013-P</json:string></doi><id>2C08A6F9965E87A5FA4832C61259974CFEA9F208</id><fulltext><json:item><original>true</original><mimetype>application/pdf</mimetype><extension>pdf</extension><uri>https://api.istex.fr/document/2C08A6F9965E87A5FA4832C61259974CFEA9F208/fulltext/pdf</uri></json:item><json:item><original>true</original><mimetype>text/plain</mimetype><extension>txt</extension><uri>https://api.istex.fr/document/2C08A6F9965E87A5FA4832C61259974CFEA9F208/fulltext/txt</uri></json:item><json:item><original>false</original><mimetype>application/zip</mimetype><extension>zip</extension><uri>https://api.istex.fr/document/2C08A6F9965E87A5FA4832C61259974CFEA9F208/fulltext/zip</uri></json:item><istex:fulltextTEI uri="https://api.istex.fr/document/2C08A6F9965E87A5FA4832C61259974CFEA9F208/fulltext/tei"><teiHeader><fileDesc><titleStmt><title level="a">Off-line cursive handwriting recognition using hidden markov models</title></titleStmt><publicationStmt><authority>ISTEX</authority><publisher>ELSEVIER</publisher><availability><p>ELSEVIER</p></availability><date>1995</date></publicationStmt><sourceDesc><biblStruct type="inbook"><analytic><title level="a">Off-line cursive handwriting recognition using hidden markov models</title><author><persName><forename type="first">H.</forename><surname>Bunke</surname></persName><affiliation>Author to whom correspondence should be addressed.</affiliation><affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</affiliation></author><author><persName><forename type="first">M.</forename><surname>Roth</surname></persName><affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</affiliation></author><author><persName><forename type="first">E.G.</forename><surname>Schukat-Talamazzini</surname></persName><affiliation>Lehrstuhl für Informatik 5 (Mustererkennung), Friedrich-Alexander-Universität Erlangen-Nürnberg, Martensstrasse 3, D-91056 Erlangen, Germany</affiliation></author></analytic><monogr><title level="j">Pattern Recognition</title><title level="j" type="abbrev">PR</title><idno type="pISSN">0031-3203</idno><idno type="PII">S0031-3203(00)X0011-8</idno><imprint><publisher>ELSEVIER</publisher><date type="published" when="1994"></date><biblScope unit="volume">28</biblScope><biblScope unit="issue">9</biblScope><biblScope unit="page" from="1399">1399</biblScope><biblScope unit="page" to="1413">1413</biblScope></imprint></monogr><idno type="istex">2C08A6F9965E87A5FA4832C61259974CFEA9F208</idno><idno type="DOI">10.1016/0031-3203(95)00013-P</idno><idno type="PII">0031-3203(95)00013-P</idno></biblStruct></sourceDesc></fileDesc><profileDesc><creation><date>1995</date></creation><langUsage><language ident="en">en</language></langUsage><abstract xml:lang="en"><p>A method for the off-line recognition of cursive handwriting based on hidden Markov models (HMMs) is described. The features used in the HMMs are based on the arcs of skeleton graphs of the words to be recognized. An algorithm is applied to the skeleton graph of a word that extracts the edges in a particular order. Given the sequence of edges extracted from the skeleton graph, each edge is transformed into a 10-dimensional feature vector. The features represent information about the location of an edge relative to the four reference lines, its curvature and the degree of the nodes incident to the considered edge. The linear model was adopted as basic HMM topology. Each letter of the alphabet is represented by a linear HMM. Given a dictionary of fixed size, an HMM for each dictionary word is built by sequential concatenation of the HMMs representing the individual letters of the word. Training of the HMMs is done by means of the Baum-Welch algorithm, while the Viterbi algorithm is used for recognition. An average correct recognition rate of over 98% on the word level has been achieved in experiments with cooperative writers using two dictionaries of I50 words each.</p></abstract><textClass><keywords scheme="keyword"><list><head>Keywords</head><item><term>Optical character recognition</term></item><item><term>Cursive script recognition</term></item><item><term>Off-line recognition</term></item><item><term>Hidden Markov model</term></item><item><term>Skeleton graphs</term></item></list></keywords></textClass></profileDesc><revisionDesc><change when="1994-06-21">Received</change><change when="1995-01-12">Modified</change><change when="1994">Published</change></revisionDesc></teiHeader></istex:fulltextTEI></fulltext><metadata><istex:metadataXml wicri:clean="Elsevier, elements deleted: tail"><istex:xmlDeclaration>version="1.0" encoding="utf-8"</istex:xmlDeclaration><istex:docType PUBLIC="-//ES//DTD journal article DTD version 4.5.2//EN//XML" URI="art452.dtd" name="istex:docType"></istex:docType><istex:document><converted-article version="4.5.2" docsubtype="fla"><item-info><jid>PR</jid><aid>9500013P</aid><ce:pii>0031-3203(95)00013-P</ce:pii><ce:doi>10.1016/0031-3203(95)00013-P</ce:doi><ce:copyright type="unknown" year="1995"></ce:copyright></item-info><head><ce:title>Off-line cursive handwriting recognition using hidden markov models</ce:title><ce:author-group><ce:author><ce:given-name>H.</ce:given-name><ce:surname>Bunke</ce:surname><ce:cross-ref refid="COR1"><ce:sup>†</ce:sup></ce:cross-ref><ce:cross-ref refid="AFF1"><ce:sup>∗</ce:sup></ce:cross-ref></ce:author><ce:author><ce:given-name>M.</ce:given-name><ce:surname>Roth</ce:surname><ce:cross-ref refid="AFF1"><ce:sup>∗</ce:sup></ce:cross-ref></ce:author><ce:author><ce:given-name>E.G.</ce:given-name><ce:surname>Schukat-Talamazzini</ce:surname><ce:cross-ref refid="AFF2"><ce:sup>‡</ce:sup></ce:cross-ref></ce:author><ce:affiliation id="AFF1"><ce:label>∗</ce:label><ce:textfn>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</ce:textfn></ce:affiliation><ce:affiliation id="AFF2"><ce:label>‡</ce:label><ce:textfn>Lehrstuhl für Informatik 5 (Mustererkennung), Friedrich-Alexander-Universität Erlangen-Nürnberg, Martensstrasse 3, D-91056 Erlangen, Germany</ce:textfn></ce:affiliation><ce:correspondence id="COR1"><ce:label>†</ce:label><ce:text>Author to whom correspondence should be addressed.</ce:text></ce:correspondence></ce:author-group><ce:date-received day="21" month="6" year="1994"></ce:date-received><ce:date-revised day="12" month="1" year="1995"></ce:date-revised><ce:abstract><ce:section-title>Abstract</ce:section-title><ce:abstract-sec><ce:simple-para>A method for the off-line recognition of cursive handwriting based on hidden Markov models (HMMs) is described. The features used in the HMMs are based on the arcs of skeleton graphs of the words to be recognized. An algorithm is applied to the skeleton graph of a word that extracts the edges in a particular order. Given the sequence of edges extracted from the skeleton graph, each edge is transformed into a 10-dimensional feature vector. The features represent information about the location of an edge relative to the four reference lines, its curvature and the degree of the nodes incident to the considered edge. The linear model was adopted as basic HMM topology. Each letter of the alphabet is represented by a linear HMM. Given a dictionary of fixed size, an HMM for each dictionary word is built by sequential concatenation of the HMMs representing the individual letters of the word. Training of the HMMs is done by means of the Baum-Welch algorithm, while the Viterbi algorithm is used for recognition. An average correct recognition rate of over 98% on the word level has been achieved in experiments with cooperative writers using two dictionaries of I50 words each.</ce:simple-para></ce:abstract-sec></ce:abstract><ce:keywords><ce:section-title>Keywords</ce:section-title><ce:keyword><ce:text>Optical character recognition</ce:text></ce:keyword><ce:keyword><ce:text>Cursive script recognition</ce:text></ce:keyword><ce:keyword><ce:text>Off-line recognition</ce:text></ce:keyword><ce:keyword><ce:text>Hidden Markov model</ce:text></ce:keyword><ce:keyword><ce:text>Skeleton graphs</ce:text></ce:keyword></ce:keywords></head></converted-article></istex:document></istex:metadataXml><mods version="3.6"><titleInfo><title>Off-line cursive handwriting recognition using hidden markov models</title></titleInfo><titleInfo type="alternative" contentType="CDATA"><title>Off-line cursive handwriting recognition using hidden markov models</title></titleInfo><name type="personal"><namePart type="given">H.</namePart><namePart type="family">Bunke</namePart><affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</affiliation><description>Author to whom correspondence should be addressed.</description><role><roleTerm type="text">author</roleTerm></role></name><name type="personal"><namePart type="given">M.</namePart><namePart type="family">Roth</namePart><affiliation>Institut für Informatik and angewandte Mathematik, Universität Bern, Länggassstrasse 5I, CH-3012 Bern, Switzerland</affiliation><role><roleTerm type="text">author</roleTerm></role></name><name type="personal"><namePart type="given">E.G.</namePart><namePart type="family">Schukat-Talamazzini</namePart><affiliation>Lehrstuhl für Informatik 5 (Mustererkennung), Friedrich-Alexander-Universität Erlangen-Nürnberg, Martensstrasse 3, D-91056 Erlangen, Germany</affiliation><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">1994</dateIssued><dateCaptured encoding="w3cdtf">1994-06-21</dateCaptured><dateModified encoding="w3cdtf">1995-01-12</dateModified><copyrightDate encoding="w3cdtf">1995</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">A method for the off-line recognition of cursive handwriting based on hidden Markov models (HMMs) is described. The features used in the HMMs are based on the arcs of skeleton graphs of the words to be recognized. An algorithm is applied to the skeleton graph of a word that extracts the edges in a particular order. Given the sequence of edges extracted from the skeleton graph, each edge is transformed into a 10-dimensional feature vector. The features represent information about the location of an edge relative to the four reference lines, its curvature and the degree of the nodes incident to the considered edge. The linear model was adopted as basic HMM topology. Each letter of the alphabet is represented by a linear HMM. Given a dictionary of fixed size, an HMM for each dictionary word is built by sequential concatenation of the HMMs representing the individual letters of the word. Training of the HMMs is done by means of the Baum-Welch algorithm, while the Viterbi algorithm is used for recognition. An average correct recognition rate of over 98% on the word level has been achieved in experiments with cooperative writers using two dictionaries of I50 words each.</abstract><subject><genre>Keywords</genre><topic>Optical character recognition</topic><topic>Cursive script recognition</topic><topic>Off-line recognition</topic><topic>Hidden Markov model</topic><topic>Skeleton graphs</topic></subject><relatedItem type="host"><titleInfo><title>Pattern Recognition</title></titleInfo><titleInfo type="abbreviated"><title>PR</title></titleInfo><genre type="Journal">journal</genre><originInfo><dateIssued encoding="w3cdtf">199509</dateIssued></originInfo><identifier type="ISSN">0031-3203</identifier><identifier type="PII">S0031-3203(00)X0011-8</identifier><part><date>199509</date><detail type="volume"><number>28</number><caption>vol.</caption></detail><detail type="issue"><number>9</number><caption>no.</caption></detail><extent unit="issue pages"><start>1283</start><end>1451</end></extent><extent unit="pages"><start>1399</start><end>1413</end></extent></part></relatedItem><identifier type="istex">2C08A6F9965E87A5FA4832C61259974CFEA9F208</identifier><identifier type="DOI">10.1016/0031-3203(95)00013-P</identifier><identifier type="PII">0031-3203(95)00013-P</identifier><recordInfo><recordContentSource>ELSEVIER</recordContentSource></recordInfo></mods></metadata><enrichments><istex:catWosTEI uri="https://api.istex.fr/document/2C08A6F9965E87A5FA4832C61259974CFEA9F208/enrichments/catWos"><teiHeader><profileDesc><textClass><classCode scheme="WOS">COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE</classCode><classCode scheme="WOS">ENGINEERING, ELECTRICAL & ELECTRONIC</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 001029 | SxmlIndent | more
Ou
HfdSelect -h $EXPLOR_AREA/Data/Istex/Corpus/biblio.hfd -nk 001029 | 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:2C08A6F9965E87A5FA4832C61259974CFEA9F208
|texte= Off-line cursive handwriting recognition using hidden markov models
}}
| This area was generated with Dilib version V0.6.32. |  |