一种基于HMM的维吾尔文联机手写识别的方法  

Method of on-line Uyghur handwriting recognition using HMM

作　　者：陈晓娇[1] 哈力木拉提.买买提 

机构地区：[1]新疆大学信息科学与工程学院多语种信息技术重点实验室,乌鲁木齐830046

出　　处：《计算机工程与应用》2013年第24期175-178,237,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.60863009;No.61032008;No.61163031)

摘　　要：在维吾尔文联机手写识别过程的训练阶段,单词被切分成字母,经过特征提取和聚类形成特征向量作为模型的输入。构造出以字符为基元的隐马尔可夫模型(HMM),将其嵌入到识别字典网络中。通过基于HMM的分类识别器,最终得到识别结果。首次将消除延迟笔画、建立有延迟笔画和无延迟笔画的字典的方法应用于维吾尔文手写识别中,取得了较高的识别率。During the process of the online Uyghur handwriting recognition, a word is segmented into characters at the training stage. Passing the state of feature extraction and clustering, each character is entirely transformed into a feature vector as the input of the models. Hidden Markov Model （HMM） for modeling the Uyghur characters as primitives is embedded in a recognition lexicon network. Then through the classification identifier based on HMM and finally the results are obtained. The delayed strokes removed and the dictionaries with and without the delayed strokes constructed which are applied in the Uyghu,r handwriting recog nition are introduced firstly, and higher recognition rate is obtained.

关 键 词：维吾尔文 联机 手写识别 隐马尔可夫模型 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]