摘要
文章针对脱机手写体汉字特征约简与识别中离散化不当的问题,构建出基于实域粗糙集的脱机手写体汉字识别决策信息系统。提出了实域粗糙集中特征属性的广义重要度的概念以及空间中的广义近邻关系;设计了在广义近邻关系下基于实域粗糙集模型的特征属性约简算法,构建出基于实域粗糙集的脱机手写体汉字识别决策信息系统;采用基于变粒度仿反馈机制的智能认知模型对脱机手写体汉字识别决策信息系统进行仿反馈识别,并建立了变粒度仿反馈机制智能识别方法的评价指标体系和认知信息粒度变换规则,提出基于变粒度仿反馈机制的智能认知算法。对SCUT-IRAC HCCLIB样本库中的汉字进行了仿真实验研究,平均识别精度达到95.37%。仿真实验表明,相比于传统认知系统单向开环方式,该文提出的方法对提升脱机手写体汉字的识别效率、可识别性及正确识别率是有效可行的。
The problems of improper discretization in the off-line handwritten Chinese character feature extraction and recognition exist in the old recognition system. In order to solve it, the off-line handwritten Chinese character recognition decision information system based on real rough set is established. Firstly, the conception of general important degree of feature attribute in the real rough set and the general neighborhood relationship in the Euclidean space are introduced. Then an algorithm of feature attribute reduction based on real rough set model is designed, and an off-line handwritten Chinese character recognition decision information system based on real rough set is constructed. Finally, an intelligent cognitive model based on variable granularity feedback mechanism is used to recognize the off-line handwritten Chinese character recognition decision information system by simulated feedback, an evaluation index system and a granularity transformation rule of cognitive information for the intelligent recognition method of variable granularity feedback mechanism are established, and an intelligent cognitive algorithm based on variable granularity imitation feedback mechanism is proposed. The simulation experiments of Chinese characters in the SCUT-IRAC HCCLIB sample dictionary are carried out, and the average recognition precision is 95.37%. Simulation results show that the method proposed in this paper improves the recognition efficiency, recognition rate and correct recognition rate of off-line handwritten Chinese characters, which is effective and feasible compared with the traditional one-way open-loop system.
作者
王芳元
王建平
WANG Fangyuan;WANG Jianping(School of Electric Engineering and Automation,Hefei University of Technology,Hefei 230009,China)
出处
《合肥工业大学学报(自然科学版)》
CAS
北大核心
2018年第3期333-341,共9页
Journal of Hefei University of Technology:Natural Science
基金
国家自然科学基金青年科学基金资助项目(61305029)
博士后科学基金面上资助项目(2013M541820
2013M532118)
中央高校基本科研业务费专项资金资助项目(2013HGBH0010
2013HGQC0012)
安徽省自然科学基金青年基金资助项目(1408085QF133)
关键词
手写体汉字识别
实域粗糙集
属性约简
变粒度
仿反馈
handwritten Chinese character recognition
real rough set
attribute reduction
variable granularity
simulated feedback