Chinese Character Recognition Based On The Model - Vector Conversion And Conversion Degree Function

Chinese character is the crystal of Chinese culture and now it can be recorded electronically in information society. The aim of Chinese Character Recognition is to study how to enable computer to be "literacy", which involves in many subjects such as Pattern Recognition, Al, Fuzzy Math, Information Theory and Computer Science etc.. It's a comprehensive technology and valuable both in practice and in theory in high-tech fields such as Office Automation, Chinese Information Handling and Al etc..In this paper, we introduce the fundamental viewpoints, theories and methods of Attribute Theory, present a Chinese Character Recognition method based on Pattern-vector Transformation, Qualitative Mapping (QM) and Conversion Degree Function(CDF), and discuss the relation between this method and the character abstraction in classical statistics pattern recognition method and the structure analysis in structure pattern recognition method. Because the character abstraction in classical statistics pattern recognition method can be regarded as an operation of Pattern-vector Transformation, and the structure analysis in structure pattern recognition method can be regarded as a (super) vector composed of many sub-patterns, We can point out that this method can also be thought of as an operation of Pattern-vector Transformation. Because vector recognition is summed up as a QM, this method has the strong points of the two classical methods.It is necessary to point out that the reason why Gauss qualitative CDF is used to calculate Similarly Degree is not only because this function can express the transformation degree from vector X to vector G (which is the similarly degree between X and G), but also because this function can induce an Artificial Neuron, have the nature of Qualitative Criteria, and at the same time, can avoid the hidden trouble of cos θ (which substitutes similarly with share line).Because of the above reasons, the Chinese character pattern memory which can be produced after the samples study and the course of the Chinese Character Recognition not only have the parallel distributed computing character of Artificial Neuron, but also obviously take on fuzzy character in the course of recognition and classification operation.In the process of computer's training and learning, this method can produce visible memory, and the training and learning is fast. However, Chinese Characters Recognition based on Attribute Theory is still in the stage of trial and exploration, and some shortcomings may exist. It is necessary to improve further in the near future.