手写体汉字识别问题综论  

A GENERAL DISCUSSION FOR HANDPRINTED CHINESE CHARACTER RECOGNITION

作　　者：路浩如[1] 杨源远[1] 

机构地区：[1]浙江大学

出　　处：《计算机应用与软件》1994年第2期1-8,共8页Computer Applications and Software

摘　　要：本文在讨论了人对汉字的认知心理试验研究结果之后,认为人类的汉字识别是经由整字属性及分层结构的多种途径复合而成的。这些途径中的大多数都可以在手写体汉字的机器识别中模仿运用。由此可把汉字的机器识别方法归纳为整模式法和子模式法两个大类,对它们的有利方面和困难方面进行了讨论;对影响其性能的因素进行了分析;对识别率以及它与识别字集大小和后处理等的关系提出了评估的问题。Following the discussion of some experimental results in cognitive psychology, the present paper considers that the Chinese character recognition of human beings is a compound of various routines through the global character attributes and the character structure from its hierarchical components. Most of such routines can be imitated in machine recognition of handprinted Chinese characters. The advantages and difficulties of global-pattern approaches and sub-pattern approaches in machine recognition are discussed and the main factors affecting their performance are analyzed. A general evaluation for recognition rate and its relation to character set size and postprocessing is proposed.

关 键 词：汉字识别 手写体汉字 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]