摘要
东巴象形文字被国际学界认为是当今世界上唯一还在使用的象形文字,用象形文字书写的经典,称为东巴经。东巴象形文字多具图画特性,结构复杂形式多样且笔划各异。针对东巴文特有的结构进行了深入的研究,主要讨论了东巴象形文的特征提取和文字识别。特征提取是文字识别中很重要的环节,文字识别中特征提取的方法有很多,但由于东巴文字的字型有很多种特点,提出了适合东巴文识别的最优特征提取方法:特征点法、投影法;识别方法:高阶神经网络法。通过实验对该方法进行了验证,结果表明该方法的可行性。
Dongba hieroglyphs by international scholars believe is only used in the world today with hieroglyphics,pictographic writing classic,known as dongba script. It has the picture characteristics and the complex structure of different forms strokes. According to the structure characteristic of Dongba we do some in-depth study. Feature extraction and character recognition of Dongba Pictograph are mainly discussed. Feature extraction is a very important link in the character recognition. there are many methods in feature extraction,but due to the many kinds of characteristics of Dongba font proposed the optimal feature extraction method for Dongba identification： the characteristic point method,projection method; identification method： high order neural network method. The method has been verified by experimental. The results show that the method is feasible.
出处
《电子测量与仪器学报》
CSCD
北大核心
2016年第11期1774-1779,共6页
Journal of Electronic Measurement and Instrumentation
基金
国家社科基金重大项目(12&ZD234)
现代测控技术教育部重点实验室开放项目(KF20161123201
KF20161123202)资助
关键词
特征提取
文字识别
东巴象形文字
神经网络
Feature extraction
character recognition
Dongba hieroglyphs
neural network
