藏文联机手写识别的研究与实现  

Research and implementation of Tibetan online handwritten recognition

作　　者：索南尖措[1] 关白[1] 李雷[1] 山发富 

机构地区：[1]西藏大学藏文信息技术研究中心,西藏拉萨850000

出　　处：《计算机时代》2017年第7期10-12,16,共4页Computer Era

基　　金：西藏自治区自然科学基金(2016ZR-15-5);西藏自治区自然科学基金(2016ZR-TU-07);西藏自治区高校青年教师创新支持计划(QCZ2016-11);西藏自治区高校青年教师创新支持计划(QCZ2016-13)

摘　　要：提出了一种基于笔画的藏文联机手写识别的方法。首先将每个藏文字分解为各个子模式,然后将子模式进一步划分为笔画,最后通过识别出的笔画和每个藏文字之间对应规则判断出使用者所书写的藏文字。识别笔画时,将其轨迹看成点的集合,使用八个方向码对其进行标记,进而确定这个笔画。在识别之前,先将相似的笔画进行归类合并,以解决使用者书写不规范的情况。在不考虑连笔的情况下,对常用的600多个藏文字进行测试,准确率可以达到92%。This paper presents a method of Tibetan online handwriting recognition based on stroke and its corresponding rules, and implements it. First, each Tibetan text is decomposed into various sub patterns, and then the sub pattern is further divided into strokes. Through the identification of strokes and the corresponding rules of each Tibetan character, the user＇s handwriting character is recognized. In the identification of a stroke, its trajectory is taken as a set of points, and eight directional codes are used to mark it, and then the stroke is deten^nined. Before the identification, the similar strokes are classified and merged to solve the problem of nonstandard user writing. Without considering the joined-up writing, tested on more than 600 Tibetan words commonly used, the accuracy rate can reach 92%.

关 键 词：藏文联机手写识别 笔画识别 方向码 笔画合并 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]