Research On Intelligent Cognitive Method Of Off-line Handwritten Chinese Characters Based On Variable Precision Feedback Structure

Off-line handwritten Chinese character recognition is one of the hot topics in the field of pattern recognition because of its large number of similar characters,many characters and variation in writing.In this paper,the handwritten Chinese characters of cognitive decision information system is constructed based on the variable precision rough set model,the relationship between cognitive accuracy and the classification results is researched,a feedback intelligent cognitive mechanism with variable precision is proposed,in order to improve the handwritten Chinese characters cognition accuracy.The main work of this dissertation is as follows:1.The method of constructing a cognitive decision information system for off-line handwritten Chinese characters is explored.Various patterns Chinese characters is analyzed,a finite field and supervised learning methods of offline handwritten Chinese characters of cognitive decision information system is constructed based on variable precision rough set theory,the knowledge sufficient criterion in the combination of multi feature information is given,the handwritten Chinese characters cognition decision information system with full information representation under finite field.2.A method based on variable precision rough set is established to reduce the feature attribute of off-line handwritten Chinese characters.A feature classification capability index based on information importance and approximate dependency degree is defined,and a Chinese character attribute reduction algorithm based on the classification ability of the feature is given,and the simplified feature set with different cognitive accuracy is obtained.3.A variable precision feedback cognitive mechanism for off-line handwritten Chinese characters is proposed.The evaluation index of the Chinese characters cognition results based on the entropy measure is established,the error entropy of the Chinese characters cognition results is defined and the transformation rule of accuracy based on the results of the evaluation is given to obtain optimal simple features of off-line handwritten Chinese characters in the cognitive process in order to reduce the complexity and improve the cognitive accuracy of the cognitive system.4.A fusion method for cognitive rules of off-line handwritten Chinese characters is explored.In view of the uncertainty in off-line handwritten Chinese character recognition process,a classifier with two levels of classifiers is designed to fuse the cognitive rules to get the real attributes of Chinese characters and reduce the rejection rate.In order to verify the feasibility and effectiveness of the proposed method,the handwritten Chinese character image database in GB23122-80 Chinese character library is used for validation.The Matlab simulation shows that this method effectively improves the recognition accuracy of off-line handwritten Chinese characters.