基于假设检验的手写印刷体汉字识别方法  

Handprinted Chinese Characters Recognition Method based on Hypothesis Testing

作　　者：黄铁军[1] 胡家忠[1] 

机构地区：[1]华中理工大学图象所,武汉工业大学计算机系

出　　处：《中文信息学报》1996年第4期20-23,共4页Journal of Chinese Information Processing

基　　金："863"计划智能计算机主题资助

摘　　要：为将统计决策方法和句法方法有机结合起来，本文提出了以部件为基元的基于假设检验的手写印刷体汉字识别方法：由统计方法得到候补字集，利用部件特征的先验知识抽取待识字可能包含的部件并对假设进行验证，从而不断缩小候补字集，并逐步完善汉字的结构描述。初步实验表明其分类效果明显。o combine statistical decision method and syntactic structural method, this paper advanced an idea about a leading syntactic structural method based on hypothesis testing that regards elements as basis units. This method refers to the classification result of decision method as the candidate set, then tries to take out possible elements by their structural features and tests the hypothesis. The process will reduce the candidate set step by step and give more and more comprehensive description of the character. The primary experiments show that the candidate set can be reduced 1/3 or more.

关 键 词：汉字识别 假设检验 手写体 印刷体 模式识别 

分 类 号：TG391.4[金属学及工艺—金属压力加工]