中文摘要：

针对手写相似汉字识别问题,提出一种新的基于分类器的方法来更全面地利用原始特征中的判别信息.新方法与现有的对相似汉字提取额外特征的方法不同,其在得到特征向量之后,首先利用修正的二次判别函数进行分类,然后用支持向量机对分类结果中的相似汉字的上述特征向量进行再一次的分类,得到最终的识别结果.利用分类混淆矩阵自动得到相似汉字集合,并提出了一种新型的存储结构用于快速查找支持向量机的训练字典.在ETL9B手写汉字数据库上的实验表明,所提出的方法可得到相对于提取额外特征方法更好的识别结果,以此证明了原始特征中存在对于相似字的判别信息,提出的基于分类器的方法可更好地利用这些判别信息.

英文摘要：

To solve the similar handwritten character recognition problem,a novel scheme is proposed to make better use of the feature＇s discriminative information.Different from the methods for extracting the extra feature for the similar characters,the Modified Quadratic Discriminant Function（MQDF）is first adopted to classify the feature,then the Support Vector Machine（SVM）is used to discriminate the similar characters without the extra feature.To collect the subset of similar characters,the confusion matrix is employed.A new structure for storing the dictionary of the SVM is also proposed for quickly searching.Experimental results on ETL9 Bshow the superior performance of the proposed scheme to the methods for extracting the extra feature,which proves that the feature contains discriminative information for the similar characters and that the proposed scheme can utilize this information very effectively.