Research On Chinese Character Image Recognition Method Based On Rough-Set

As classification problem of multi-dimension pattern space, off-line handwritten Chinese character recognition is one hot research in pattern recognition field. Rough-Set theory is one soft computing method treating the data without precision and certainty and completeness. It is built on a base of classification mechanism. Classification is comprehended by it as equivalence relation, and knowledge is comprehended as partition of data. Rough-Set theory can treat uncertain problem relative objectively. It can get a biggish theory meaning and practice value that Rough-Set is used for the off-line handwriting Chinese characters recognition.The following are the major contents in the thesis:1. The construct of handwritten Chinese character decision information system. Based on the concept of information system in generalized domain defined by Rough-Set, the handwritten Chinese character information system is defined by the paper. Imitating human recognize character by supervised training, handwritten Chinese character decision information system is constructed by adding real attribute of Chinese character samples in handwritten Chinese character information system.2. Characteristic quantity reduction. The essentiality of handwritten Chinese character attribute is figured by entropy value. One handwritten Chinese character characteristic quantity reduction algorithm is designed to decrease the redundancy degree of attributes in recognition.3. Redundancy degree reduction of multi-sample Chinese character. Based on rule conflict judgment, one heuristic redundant samples reduction method is suggested to reduce the redundancy degree of multi-sample Chinese character.4. Recognition of multi-font Chinese character. The Chinese character set with multi-template and multi-attribute is constructed. One ruler fusion method for Chinese character recognition based on confidence is presented. One handwritten Chinese character recognition system is designed.5. The SCUT-IRAC handwritten Chinese character sample database is selected to verify the effectiveness of the attributes reduction method, the samples reduction method, and the ruler fusion method.In the condition of MATLAB6.5, The SCUT-IRAC handwritten Chinese character sample database is selected as experimental samples for simulation experimental verification. The results of experiment indicate : the attributes reduction method is available for the reduction of multi-dimension statistical character attribute; the heuristic redundant samples reduction method is effectual for the reduction of the redundancy degree of multi-samples; the ruler fusion method based on confidence can elevate the recognition accuracy rate of handwritten Chinese character.