上下文感知手写数学公式结构分析  

Context-Aware Handwritten Math Formula Structural Analysis

作　　者：任昆[1] 戴永[1] 王求真[1] 李璇[1] 喻世东[1] 

机构地区：[1]湘潭大学智能计算与信息处理教育部重点实验室,湖南湘潭411105

出　　处：《湘潭大学自然科学学报》2014年第2期85-91,共7页Natural Science Journal of Xiangtan University

基　　金：湖南省教育厅项目(13C914);湖南省十二五重点学科建设项目

摘　　要：不同教学层次数学公式子结构的使用概率存在明显差异.该文结合符号类别及数学公式子结构在不同教学层次的教学概率,提出基于上下文感知联机手写数学公式符号区位跟踪结构分析方法.将数学公式书写结构分为一般和非一般两大类,依据教学概率分别构建区位上下文感知模型,利用模型进行后继感知区位监测、区位推荐、符号的区位确定.实验表明,区位确定的正确率达到98.2%;区位确定的平均时间为3ms左右,与已有成果比较明显提高,有效加快了手写数学公式识别速度.With regard to different teaching levels,there is obvious difference in the usage probability of the mathematical formula substructure.Combined symbol type and teaching probability of substructure of mathematics formula used in various teaching levels,this article puts forward symbolic location tracking （structural analysis） method of online handwritten mathematics based on context awareness in different teaching levels＇s teaching probability.Formula writing process structure is divided into general and non-general types.Context awareness model is designed in line with teaching probability.The model is used to carry out subsequent perception location monitoring,location recommendation as well as symbol location determination.Experiments show an accuracy rate of 98.2％ and an average time of 3 ms in location determination,making great improvements on existing results and effectively accelerating the speed in handwritten math formula recognition.

关 键 词：上下文感知 手写数学公式 区位跟踪 模型 教学概率 结构分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]