Based On Qualitative Mapping And Conversion Degree Function Of Chinese Character Recognition

Character recognition is an important branch of Pattern Recognition whose duty is to enable the computer to be "literacy". It is an integration technology involving Pattern Recognition, image manipulation, digital signals process, fuzzy math, etc. And it is very valuable both in practice and theory in high tech fields such as Office Automation, machine translation, Al and Chinese information processing.After introduced the foundational viewpoints, methods and theories in Attribute Theory, we developed a brief Chinese characters recognition system using Qualitative Mapping(QM) and Conversion Degree Function(CDF) and expatiate the method in this paper. We look a character (pattern) as the set of points it contains and get a Memory Pattern of this character from several samples. The similarity degree of the input and reference pattern is get with mWCDF. The introduce of CDF brings fuzzy property to the course of machine learning and recognition, which exactly reflects the fuzzy property of the thinking course of human brain.The main innovation of this method is the actual Memory Pattern. And unlike the statistical method, we don't need many samples(only several is enough). The result of this Chinese characters recognition program proves our method can learn a pattern vary rapidly (spending only 15 seconds to training 1280 Chinese characters). The recognition rate of learned printing character is 100% and of handwriting character is 86.2%.Li Wenpei (Computer Software and Theory) Directed by Prof. Feng Jiali...