A Recognition Research Of Off-line Handwritten WuMei Tibetan Based On BP Network

So-called off-line recognition,it is also called the Optical Character Recognition, and it's abbreviation is OCR,Handwritten or printed document will be transform into the image by the scanner,then will be recognition into printed characters by the computer.The main task of handwritten character recognition is pattern classification, the recognition method is extracting the features from handwritten image and using the features to classify the characters in recognition dictionary.Because the styles of handwritten characters are different from person to person,even the same person who writes the characters will be some changes,and will be unable to obtain real-time information,the characteristic will enable it to become the most difficult branch of handwritten recognition domain.WuMei Tibetan,as a language which is popular in Tibetan area,is of great significance to study its handwritten recognition methods of promoting Minority nationality areas information technology development.Tibetan writing is one kind of alphabetic writing which is different from Chinese and English.,its written way is unique in the world.Tibetan is crosswise writen from left to right,and the word is composed of the characters,each character is constituted by strokes,and it is separated by a Separation sign between word and word.The stroke of WuMei Tibetan is simple,and the written speed is quick,pitch is big between word and word,row and row.Some superposition letters distort completely,specially WuMei Tibetan writing style diversification,the different mode of writing also has its unique characteristic.These characteristics make Handwritten WuMei Tibetan recognition very difficulty.The paper mainly discussed the recognition techniques of off-line handwritten Wumei font style.After careful study of popular character recognition techniques and the features of Wumei font style.We proposed a technique of off-line handwritten Wumei font style,that is a Tibetan recognition plan based on the BP neural network classification.And we've developed a system which could recognize handwritten Wumei font style.The recognition rate of consonant letter of certain person is 95%. The system has a stable performance and it could be used in practice. The test result proves that the technique we've employed is feasible and it is a base to extend its functions in future research.Of course,there are still some problems need to be dissolved such as the display of Tibetan characters,the classification of similar characters,and low recognition rate etc.We will resolve these problems step by step in the further study and transform the experiment system into suitable product..