候选字静态生成技术及其在两级LDA汉字识别中的应用(AStatic.PDFVIP

Proceedings of the 26th Chinese Control Conference July 26-31, 2007, Zhangji aj ie, Hunan, China 选字静态生成技术及其在两级LDA 汉字识别中的应用* 刘志斌，金连文+ 南理工大学电子与信息学院，广东, 广州 510641 +Email: lianwen.jin@ 要：线性判决分析 (Linear Discriminate Analysis, LDA)作为一种有效的特征选择工具已经被广泛地运用在汉字识 别当中。本文在传统LDA 方法基础上，提出了一种新颖的候选字静态生成技术，从而克服了LDA 训练时间长、计算 量大的缺陷，实现了一种基于两级LDA 的大类别手写汉字分类方法。实验表明， 用两级LDA 分类方法与单纯 用 最小距离分类器和 用了一级LDA 的分类方法比较，误识别率分别大幅降低了60%和 35% 。 关键词：线性判决分析,手写汉字识别,相似汉字 A Static Candidates Generation Technique and its Application * in Two-stage LDA Chinese Character Recognition Liu Zhibin, Jin Lianwen+ College of Electronic Information, South China University of Technology, Guangzhou 510641 Guangdong, P. R. China +Email: lianwen.jin@ Abstract: As an effective tool for feature selection, Liner Discriminate Analysis (LDA) has been widely used in the field of Chinese Character Recognition. In this paper, we propose a novel static candidates generation technique, which significantly reduces the storage and the computational complexity of the traditional LDA. Using the proposed technique, a two-stage LDA recognition scheme for Chinese character recognition is presented. Compared with minimum distance classifier and LDA plus minimum distance classifier, the error ratio of proposed scheme significantly decline 60% and 35% respectively, which shows the validity of the proposed approach. Key Words: Linear Discriminate Analysis ，Handwritten Chinese Character Recognition ，Similar Chinese Characters 1 引言(Introduction)1 集合q2 LL经过若干次选择和识别后，能较可靠地把 待识别字与其他汉字区分出来。 文字识别是模式识别的一个重要分支，是人工智