首页> 外国专利> ON-LINE LEARNING SYSTEM FOR CHARACTER RECOGNITION SYSTEM BASED ON NEURAL NETWORK

ON-LINE LEARNING SYSTEM FOR CHARACTER RECOGNITION SYSTEM BASED ON NEURAL NETWORK

机译:基于神经网络的字符识别系统在线学习系统

摘要

PROBLEM TO BE SOLVED: To decrease the sorting errors by preparing a secondary sort as a part of a neural network to temporarily readjust a sorting border line between the recognition of characters and the recognition of another medium. SOLUTION: A secondary sort 15 is prepared in a computer system 13 as a part of a neural network 12 including a primary sort. The training characters are first learnt by an artificial neural network system and stored in the primary sort to be recorded there. The characters which are recognized by a character recognition system with high reliability and related with the training characters are learnt again by the artificial neural network system and then stored in the sort 15 to be recorded there. Thus the sorting errors are decreased by readjusting temporarily a sorting border line between the recognition of characters and the recognition of another medium.
机译:解决的问题:通过将辅助排序作为神经网络的一部分来准备,以临时重新调整字符识别和另一种介质的识别之间的排序边界线,从而减少排序错误。解决方案:在计算机系统13中准备了次级分类15,作为包含初级分类的神经网络12的一部分。首先通过人工神经网络系统学习训练角色,并将其存储在主要类别中以在此处记录。由字符识别系统识别的具有高可靠性并与训练字符相关的字符由人工神经网络系统再次学习,然后存储在类别15中以在那里记录。因此,通过临时重新调整字符的识别和另一种介质的识别之间的排序边界线来减少排序错误。

著录项

  • 公开/公告号JPH10254996A

    专利类型

  • 公开/公告日1998-09-25

    原文格式PDF

  • 申请/专利权人 XEROX CORP;

    申请/专利号JP19980042684

  • 申请日1998-02-24

  • 分类号G06K9/66;G06F15/18;

  • 国家 JP

  • 入库时间 2022-08-22 03:07:59

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号