模糊集在手写数字识别中的应用  

The Application of Fuzzy Set to the Recognition of Handwritten Digits

作　　者：朱学芳[1] 石青云[1] 程民德[1] 

机构地区：[1]南京大学声学所,北京大学信息科学中心

出　　处：《模糊系统与数学》1996年第2期70-75,共6页Fuzzy Systems and Mathematics

摘　　要：为提高识别速度，本文提出先对文字图象进行直（折）线段、圆的ＬＭＳＥ近似并用Ｆｕｚｚｙ定义算法进行描述，以得到对书写工整的部分数字的快速识别，其速度提高近两倍（达到３７５个／秒），约有２０％的字在这一过程中就可以被识别。本文叙述的算法已经用在我们进行的无限制手写数字识别系统的实验中；反复使用表明：本文提出的方法既快又准确。In order to obtain rapid recognition we use Fuzzy label algorithm (FLA) for the recognition of handwritten digit in this paper. First, we describe some of well-written digits by means of circle approximation or line segment approximation by means of LMSE method, then. And we use FLA and other experimental parameters to reason,therefore final judgements are obtained. It is very convienent to the discrimination of well-written digit patterns. The proposed method has been used in the first stage of our OCR system. Experiments show that it works effectively and almost doubled the recognition speed (375 digits per second on EDC-5000 wrokstaion). About twenty percent of input patterns are recogonised by this Fuzzy set method after test on a lager database.

关 键 词：模式识别 模糊集 数字识别 手写数字识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] O159[自动化与计算机技术—计算机科学与技术]