基于神经网络的印刷体数字字符的识别

Printed numeric character recognigion based on neural network

提出了一种基于BP网络的印刷体数字字符的识别方法。通过对BP网络的研究与学习,设计了一种结构合理,收敛速率快的BP网络。实验结果表明,该方法对标准的印刷体数字字符的识别率达到了100%,对有1～3度倾斜角度的字符识别率也达到了96%以上。

Amethod based on BP neural network was put forward to realize printed numeric character diagnosis, with satisfied results in experimental test. The rate reaches 100% for standard printed numbers and more than 96% for tilt character with 1-3 angles.