期刊名称：International Journal of Advanced Research In Computer Science and Software Engineering
印刷版ISSN：2277-6451
电子版ISSN：2277-128X
出版年度：2012
卷号：2
期号：5
出版社：S.S. Mishra
摘要：Od ia is one of t he sc hed uled lang uag e of Indi a. It is p rinci pal communi catio n in the stat e of Odisha, sp oken by over 32 millio n people co mprising 86% of populatio n (2 001 cencus). Optica l cha ract er recognition (OCR) is a typ e of document ima ge a nalysi s where scanned d igital ima ge that contains either machine printed or ha ndwritten script input into an OCR so ftware eng ine and translating it into an editab le machi ne reada ble d igital tex t format. In this pa per we d esigned a novel and ro bust hyb rid recog nitio n system for Odia ha ndwritten cha racter as well as we designed the st anda rd d eviation & a nd zo ne centroid based fea ture ext racti on metho d fo r better accuracy whi le tra ining and testing the Neural Network . The OHCR System is ba sed on the a lgorithm of feed fo rwa rd BPNN co mbined with GA t o perform the opti mum fea ture extractio n and recognition. The Odi a character sets a re classified into fo ur g roup s accord ing t o similarity o f their shapes a nd fea tures. The system uses fi ve ANN ha ving different pa rameters, and the output of these five ANN are fed into GA whi ch chooses the fit test and the best solution and provid e us the recog nized cha ract er.
关键词：Artificial Neura l Network; Geneti c Alg orithm; Character Recogni tion; Mo rpho logical analysis; Stand ard ;deviati on.
