基于HMM的分类器在联机手写藏文识别中的应用  

Application of Hidden Markov Model in On-line Recognition of Handwritten Tibetan Characters

作　　者：梁弼[1] 王维兰[1] 钱建军[1] 

机构地区：[1]西北民族大学中国民族信息技术研究院,甘肃兰州730030

出　　处：《微电子学与计算机》2009年第4期98-101,104,共5页Microelectronics & Computer

基　　金：国家自然科学基金项目(60273090)

摘　　要：为了解决联机手写藏文识别中藏文的曲线型笔划比较多,连笔情况很普遍以及相似字丁多等问题,提出了一种新的联机手写藏文识别方法:基于HMM分类器的联机手写藏文识别的方法.设计了三种不同的HMM分类器进行藏文字丁识别,实验结果表明,基于HMM分类器的联机手写藏文识别具有较高地识别率,前十位识别率可达93.9012%.In order to solve Tibetan curve strokes to be quite many, is very common including the situation where the Ti- betan character is written in a nonstop manner and many similar Tibetan characters in on-line recognition of handwritten Tibetan characters, we proposed a new method of on-line recognition of handwritten Tibetan characters： an on-line recog- nition of handwritten Tibetan characters method based on HMM classifiers. We designed three kinds of different HMM classifiers to distinguish Tibetan characters, the experimental results show that HMM-based classifiers on-line recognition of handwritten Tibetan characters has higher recognition rate, recognition rate of the first ten characters reach 93. 9012%.

关 键 词：联机手写藏文识别 隐马尔可夫模型 HMM分类器 识别率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]