基于神經(jīng)樹的英文字符識別技術研究.pdf

1、山東師范大學碩士學位論文基于神經(jīng)樹的英文字符識別技術研究姓名：李慶昌申請學位級別：碩士專業(yè)：管理科學與工程指導教師：劉希玉20090603山東帥范人學頒lj學位論文ABSTRACTCharactersareanimportanttooltocommunicateinfo肌ation，Howe行ectiVelyentero靠二linehandwrittencharactersintothecomputerhasbec鋤eabo砌enec_

2、kofmeprocessofcountryinfo肌ationCharactersrecognitionisanimportantresearchfieldofpattenlrecognitionCharacterrecognitiontechnologyisusedVeryextensiVelyItcanr印IacethemanandthecontmllerautomaticallyputcharactersaIldotherinfo

3、肌ationtoidentifiedbeforee11trythemintothecomputertCharacterreco印itiontechnologycansolVealargenumberofcharactersintheautomaticen臼yprobleHl，thee骶ctiveconservationofthepeopleengagedina1a唱enumberofcharacterentrytime，andgetpe

4、opleridoutthetediousworkSinceen西ishlettersarecommonlyusedalloVertheword，a11dinmany印plicationssuchasstatemellts，bills，checks，etc，itisdimcultt0be兩nted，andthereliabilityofmeidentificationishiglllydemaIldedThus，off二linehandw

5、打ttencharacterrecognitiontechnologyhasImportantpracticalvalueThecharacterrecognitionprocesscontainspre—processing，feamreextraction，identification，aIldsoonm旬orst印sBecauseofthestnIctureofneuraln℃e，ltalS0hasafeatureextracti

6、on如nctionwhentreestnJctllreisphylogeneticed，someimportantcharacteristicswillbecopiedandtransimissionasaleaVessIlbnode，eVelltuallycVolvedtobetheneuraltreeleafnodeisaIlimportantfe姍toeXtractSothispaperuseneural骶etoidenti矽fo

7、ro瑚inehandwrittencharacters，tonJrtherimprovetheemciencyofcharacterreco伊itionThisstudyaimedtodesignrecognitionmodebasedonneural仃ee，theobjectforid咖i黟isaSpecificcharactersetofE峭ishcharacters(26uppercaseletter5；ofthealphabet

8、)ThemajnrcsearchworkisorganizedaroundmehandwrittellCharacterrecognitiontechnologybasedonneuralneeFirstofall，misp印ers咖arizestherese卸chback鯽udandcurrentsituationofcharacterreco鱸itionSecondlya11a1)，Zedthepre—processingofhan

9、dw訂ttenCharactcrsimagesincludingsmooth，thimlingandsoonReseachmebasictheo叫ofneuraltreeandhowt0usedittoreco印itionh飆dwritingcharactcrsthcreseachfocuseSontreestmcture0ptimizationalgorithm鋤ditSp眥IllleterSoptimzationalgoritllI

10、llFinallysetupao昏linehalldwrittenrecognitionsystemThesyst鋤mainlycomposedbythepreprocessingmodule，feanlreextractionmoduleandtheneural1’reeidentficationmoduleThemainjobofthispapeincludefollowingaspects：(1)Setupanemal仃eeide

11、nti矽model，趾dmal【easinmlationteStinginmatlab廿Ⅳim眥entThee虢ctofitsidentifcationwerecomparedwiththeBPnetwork’sAccordingt0testresults，analysisoftheperfomaIlceofrecognitionmodelIIldicationthatthenemaltreemodeimpr0Vedthee伍cienc