论文部分内容阅读
本文评论了最近九年,特别是1970年以来的汉字识别工作。所包括的题目有:1.印刷体汉字识别。近年来所作的工作大约有半数是利用造句法描述,轮廊投影,模板匹配,付氏变换和分级处理等各种方法来解决这个问题的;2.手写体汉字识别。这是稍有些困难的工作,为解决该问题,曾尝试过笔划产生技术和统计 Markov 链分析等方法;3.联机识别。有几位研究者报告了联机汉字识別工作。联机识别的输入是动态的,输入设备有图板等合适装置。由于笔划在时间上是孤立的,其顺序是巳知的,故实质上这是一类较易完成的识别任务。4.联机分析。人们对这类问题的抱负没有对联机识别那么大。这个工作在实际中已得到应用,其中有计算机辅助教学和查字典。
This article reviews the past nine years, especially since 1970, Chinese character recognition. Topics covered include: 1. Printed Chinese character recognition. Approximately half of the work done in recent years is to solve this problem by using syntactic description, outline projection, template matching, Fourier transform and hierarchical processing. 2. Handwritten Chinese character recognition. This is a slightly harder task. To solve this problem, we have tried strokes generation techniques and statistical Markov chain analysis methods. 3. Online recognition. Several researchers reported online Chinese character recognition. The input for on-line identification is dynamic and the input device has a suitable device such as a drawing board. Since strokes are isolated in time and their order is known, it is essentially a type of easier task of identification. 4 online analysis. People’s ambitions for such issues are not as large as online. This work has been applied in practice, including computer-assisted instruction and dictionaries.