基于外接同心圆结构特征的自由手写体数字神经网络识别方法  

Neural Network Recognizer for Unconstrained Handwritten Numerals Based on Feature of Addendum Concentric Circles

作　　者：刘志敏[1] 施鹏飞[1] 郭钧锋[1] 

机构地区：[1]上海交通大学图象处理与模式识别研究所,上海200030

出　　处：《中文信息学报》1997年第2期48-54,共7页Journal of Chinese Information Processing

基　　金：国家自然科学基金

摘　　要：本文提出了一种基于外接同心圆结构提取贯穿特征码的自由手写体数字的神经网络识别方法。该方法是用自由手写体数字的外接同心圆来抽取其贯穿特征码，将获得的模式特征训练改进的BP神经网络分类器，从而达到快速分类的目的。将其应用于邮政编码识别系统，单字的识别率达到97％以上，整信的识别率可达到92％以上，得到了令人满意的结果。A neural network recognizer is proposed for unconstrained handwritten numerals based on the feature of addendum concentric circfor. The method is extracting the cross feature codes of addendum concentric circles of unconstrained handwritten numerals,using the obtained cross feature codes to train the back propagation neural network classifier,and then recognizing these characters. Applying to the zip code recognition system,we achieve that the average recognition rate of single digit is above 97%,the average recognition rate of total letter is about 92%. The test results are very satisfactory.

关 键 词：贯穿特征码 特征抽取 手写体数字 字符识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]