传统Hopfield ANN应用于汉字识别的一种改进  

THE IMPROVEMENT OF CHINESE CHARACTERS RECOGNITION USING TRADITIONAL HOPFIELD ANN

作　　者：李雪春[1] 李多[1] 马争鸣[1] 

机构地区：[1]中山大学无线电电子学系,广州510275

出　　处：《计算机应用与软件》1999年第3期26-29,25,共5页Computer Applications and Software

摘　　要：本文在传统Hopfield ANN基础上,对汉字识别某些方式提出一种新的方法,并在输出匹配的可行性上进行新的探索。在汉字字库的建立方法上提出新的观点。研究表明,本文所述方法是行之有效的,提高了汉字识别率。This paper proposes a new way on Chinese characters recognition, based on traditional Hopfield ANN. An approach to this method is made on outputmatching. Instead of using updated result directly, we set up a character database and output from this database. Using traditional Hopfield ANN only, we often obtain recognition results unclearly or unsatisfactorily. Experiments show that the way proposed here is efficient. The rate of Chinese characters recognition is improved.

关 键 词：汉字识别 人工神经网络 模式识别 汉字信息处理 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]