特征融合及相似度判据在英文识别中的应用  

Application of Feature Fusion and Similarity Classifier in English Recoghition

作　　者：吴锐[1] 赵巍[1] 尹芳[1] 唐降龙[1] 

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《计算机工程与应用》2005年第16期55-57,共3页Computer Engineering and Applications

摘　　要：英文字符识别率依然有待提高,它是进一步改善英文文档识别性能的重要内容。文章通过大量的实验,选取穿越特征和网格特征作为英文字符的分类特征,并通过实验得到一组较优的加权因子将两种特征合并,实现特征融合。基于上述混合特征,利用图像相似性测度作为分类判据,实现了一个高性能的英文字符识别系统。English Character Recognition is an important content to improve the capability of English document recognition system,and its still being improved.It has been decided by experiment that through feature and gridding feature are chosen as the classification features of the English character recognition,and feature fusion are implemented by getting a set of coefficients added to the feature vector from the experiment.According to above feature fusion,similarity classifier is used to achieve an English Character recognition system with high performance.

关 键 词：英文字符识别 特征融合 图像相似度 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]