Variable Precision Self-adaptive Feedback Adjustment Mechanism-based And Recognition Study For Handwritten Chinese Character

Handwritten Chinese characters have a large amount of words, numerous categories, complex structure, many similar words, large writing deformation, offline handwritten Chinese character image recognition is still one of pattern recognition problems. Based on variable precision rough set theory in this paper, the handwritten Chinese character recognition decision information systems is constructed. Self-adaptive feedback adjustment attribute reduction mechanism and identification methods is studied and this paper explores off-line handwritten Chinese character image recognition method to improve the recognition rate.The main work of this paper:1.A construction method of handwritten Chinese character recognition variable precision complete decision information system is explored. Based on variable precision rough set theory in this paper, the handwritten Chinese character recognition variable precision decision information systems is constructed. By analyzing the feature model of object identification, The adequacy evaluation constraints of the information capacity are established and a multi-feature fusion pattern decision information system is built by containing sufficient identifying information.2.A method of variable precision characteristics attribute reduction for handwritten Chinese characters based on information granularity entropy is designed. Based on the information entropy theory. The index for evaluating the performance of feature selection of variable precision information granularity entropy is defined in the complete information system to obtain the complete optimal reduction feature set under different precision conditions through using the method of reduction of relative information redundancy.3.A self-adaptive feedback adjustment mechanism for handwritten Chinese characters is proposed. Based on the scope of providing the roughness correction, heuristic dynamic feedback adjustment reduction mechanism for handwritten Chinese characters is constructed, the optimal reduction feature set of training samples is selected in all previous feedback process to achieve high accuracy.4.The weighted Bias recognition mechanism for offline handwritten Chinese character is established., the use of confidence maximum value determination method is based on weighted Bayesian theory by calculating Bayes confidence of each decision rule and making the threshold determination to get the real value of Chinese characters and improve the recognition accuracy.In order to verify the effectiveness of the variable precision adaptive feedback reduction and weighted Bayesian recognition which is proposed in this paper, the SCUT-IRAC handwritten Chinese characters image sample database announced from the South China University of Technology is used. Using the software algorithm that regard Rosetta and Matlab7.0 platform, The experiments use of computer software for feature extraction, attribute reduction, experimental character recognition procedure of handwritten Chinese character. The experiments show that this recognition model has been greatly improved in the intelligent direction than those previous identification systems.