具有变精度反馈机制的脱机手写体汉字智能认知  

OFFLINE HANDWRITTEN CHINESE CHARACTER INTELLIGENT RECOGNITION WITH VARIABLE PRECISION FEEDBACK MECHANISM

作　　者：王建平[1] 刘雪景 陈克琼 李帷韬[1] Wang Jianping 1, Liu Xuejing 1, Chen Keqiong 2, Li Weitao 1(1School of Electric Engineering and Automation,Hefei University of Technology,Hefei 230009,Anhui,China;2 Hefei University,Hefei 230601,Anhui,China)

机构地区：[1]合肥工业大学电气与自动化工程学院,安徽合肥230009 [2]合肥学院,安徽合肥230601

出　　处：《计算机应用与软件》2018年第7期231-236,共6页Computer Applications and Software

摘　　要：针对之前变精度粗糙集下精度β值固定的方法,面对实际样本其分类效果与β的取值之间的关系无法确定的问题,提出一种变精度反馈的脱机手写体汉字智能认知模型。定义了表征变精度粗糙集下特征属性分类能力的指标并给出约简算法。该算法中不同精度β下会得到不同维数的特征约简集,因此定义了精度β变换规则并且定义了认知结果评价机制,通过评估认知结果的可信度自适应调节精度β,实现面对不同样本能以最优精度β得到最约简特征集合进行认知。仿真实验表明,该方法汉字认知准确率可以达到92.8%。Aiming at the drawbacks of the relationship between the result of classification and the value of β cannot be determined in face of actual samples by the method of the fixed precision β,an intelligent cognitive model with variable precision feedback mechanism was proposed for the off-line handwritten Chinese character cognition. Firstly,the index of the classification of feature attributes in variable precision rough sets was defined and the reduction algorithm was given.In this algorithm,feature reduction sets of different dimensions were obtained under different precisions β. Therefore,the accuracy β transformation rules were defined and the cognitive outcome evaluation mechanism was defined. By adjusting the credibility of the cognitive results,the accuracy β was adjusted adaptively to achieve the knowledge that the different samples was obtained with the best reduction feature set with the best accuracy β. Finally,simulation experiments showed that this method achieved 92. 8% cognitive accuracy.

关 键 词：变精度 约简 智能认知 认知结果 反馈 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]