脱机手写体数字识别技术研究

脱机手写体字符识别技术是当前的热点和难点问题,是解决目前大量已有的文档资料录入工作的关键,该领域目前还没有十分成熟的产品,还正处于研究阶段,特别是如何提高其识别率已成为一个亟待解决的问题。本文根据数字笔画特征值和环的数目构建一种神经网络特征输入的新的特征提取方案,参照汉字基本笔画定义,采用环、横、竖、撇、捺、竖定义数字的五种笔画,运用单字单网的12个并行BP神经网络进行数字识别,为解决手写体字符识别中所存在的困难找到一种切实可行的新途径。

Research on Offline Handwritten Numeral Recognition Technology

Offlineh andwrittenc haracterr ecognitiont echnologyi s one of hot and difficult problems, and it is key and solution of inputting large number of documents. There are still not very mature products in the field. Especially, it is an urgent problem how to improve their recognition rate. In the paper, according to strokes eigenvalues and number of Central the new feature extraction program of neural network input are constructed. Reference on basic strokes of Chinese, the strokes of number was defined by Central,h orizontal,a cross,l eft-falling,r ight-falling. Twelve parallel digital BP neural network were applied to numeral recognition, which is a new way to solve HandwrittenN umeralR ecognition.