手写体数字识别系统中一种新的特征提取方法  

A new feature extraction method on handwritten digits recognition system

作　　者：钟乐海[1] 胡伟[2] 

机构地区：[1]西华师范大学计算机学院,南充637002 [2]绵阳师范学院计算机科学系,绵阳621000

出　　处：《四川大学学报（自然科学版）》2007年第5期1000-1004,共5页Journal of Sichuan University(Natural Science Edition)

基　　金：四川省重点科技基金(02GG006-036)

摘　　要：本文对手写体数字的特征提取方法进行了讨论,分析了当前使用较多的三种特征提取方法,指出了各自的优缺点,并在此基础上提出了一种新的特征提取方案,即13点特征提取方法,该方法从每个字符中提取关键的13个点作为特征点,其主要特点是特征提取简单有效,节省了特征提取时间,提高了识别系统的运行速度.用同一种网络对特征提取后的结果进行训练和识别.仿真结果表明,13点特征提取法用于手写体数字的识别有着极好的适应性,在运行速度和识别率上比起其它算法都有很大的提高,从而极大地证实了新算法的有效性及实用性.This paper discussed the handwritten digits feature extraction, analyzed current three kinds of feature extraction method, pointed out the advantage and weakness, and in this foundation proposed one kind of new feature extraction method, namely 13 points feature extraction method, this method withdrew the key 13 spots as characteristic point from each digits character, it is simply and effective, and saves the feature extraction time, and enhances the recognition system running rate. the experiment indicated, 13 points feature extraction method to use in the Handwritten digits recognition have the extremely good compatibility, it has the very big enhancement in the running rate and recognition, testify the new algorithms efficient and practicable.

关 键 词：手写体数字识别 特征提取 模式识别 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]