手寫體數(shù)字識別中的特征提取和特征選擇研究.pdf

1、北京郵電大學(xué)碩士學(xué)位論文手寫體數(shù)字識別中的特征提取和特征選擇研究姓名：董慧申請學(xué)位級別：碩士專業(yè)：信號與信息處理指導(dǎo)教師：盛立東20070308ABSTRACTHand而tingdigitsrecognitionisanimportamrcs∞r(nóng)chsubjectincharacterrecognitionThedi銜culti鎬ofh粕dwritingdi西tsrecognitionareduetoitsvadousaIl鋤。叩hos

2、isAtpresent，thedigitsrecognitionsofdi毹r印tlett晌ms，especiallytheomineh齟dwritingdigitsrecognition，arcstillundcrdevel叩ment，齟dt11erecogIlitiOnem娥isnotidealTherefore，itisstillaveryiInponantresearchdirectiontostudysimpleandhigh

3、e銜cienth鋤d、Ⅳritingdigitsn瑚gnitionThethesisprobesintothekeyissueofhaIld、州tingdi百tsrecognitionfeatIlreexmlction鋤dfeatureselectionThemainworkofthe也esisincludesthefollowingaspects：1Basedontheresearch囂onthefeaturesofse、remlh鋤

4、d、州tingdigits，themesisex仃actst11estructllres卸dstatisticf細rcsofsevcnkindsofhand、槲tingdi百ts，ie叫tlinefeature，s仃okedensit)rfcature，wide鰣dfeaturc，b叫centcrandbarycemerdistancefeaturc，thefirstblackpointposjtionfeature，projectf孤

5、re，andFo嘶erswitchfeature2Fmmdiffbrcntf婦tIlreselectionmcthods，tIlisthesisadoptsthreemcthodsinncr加douteraIlalogy，KWchecl【ingande劬ropymnctiont0selecttllefcatures311lismesis鋤alyzesthef’齟turedimensiondecreaseisSueofthehandw—t

6、iI培digitsthroughalotofexp耐ments4Thismesisestablishesahand、)Ir—tingdi百treco印itionsystembasedonBPneuralnetworkTheoriginalfcatulcsandselectedfcaturcsbothhavegoodsystematicperfo曲鋤ceafIcrchcckedm]rou曲BPneIlmlnetvyofI(，whichpm