并行特征融合在金融手写汉字识别中的应用  

Application of Parallel Feature Combination in Financial Handwritten Chinese Character Recognition

作　　者：温昌兵[1] 杨扬[1] 颉斌[1] 

机构地区：[1]北京科技大学信息工程学院电子信息系,北京100083

出　　处：《计算机工程》2005年第19期178-179,共2页Computer Engineering

摘　　要：针对金融票据自动识别应用中的脱机手写体汉字识别进行特征提取的研究,首先提出了用Gabor特征和Zernike矩特征来分别表征汉字的局部特征和全局特征。针对传统的串行特征融合方法的缺陷,提出了一种并行特征融合方法,将Gabor特征和Zernike矩特征组合成新的特征向量,然后使用广义K-L变换对新特征向量的维数进行压缩,去除冗余信息。实验结果验证了该方法的有效性。This paper researches on feature extraction methods for off-line handwritten Chinese character recognition applied to automatic recognition of financial bills. At first, Gabor feature and Zernike moment＇s feature are proposed to represent the partial and global feature. Considering the weaknesses of traditional serial feature fusion technique, a novel parallel features fusion method is proposed, which combines Gabor feature and Zernike moments feature to new feature vectors. And then generalized K-L transform is used to decrease the dimensions of new feature vectors so as to get rid of redundancies. Experiment result shows the effectiveness of the proposed approach.

关 键 词：特征提取 GABOR变换 ZERNIKE矩 特征融合 广义K-L变换 金融票据 脱机手写体汉字识别 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]