Neural Tree-based English Character Recognition Technology

Characters are an important tool to communicate information, How effectively enter off-line handwritten characters into the computer has became a bottleneck of the process of country information. Characters recognition is an important research field of pattern recognition. Character recognition technology is used very extensively. It can replace the man and the controller automatically put characters and other information to identified before entry them into the computer. Character recognition technology can solve a large number of characters in the automatic entry problem, the effective conservation of the people engaged in a large number of character entry time, and get people rid out the tedious work.Since english letters are commonly used all over the word,and in many applications such as statements, bills, checks, etc., it is difficult to be printed, and the reliability of the identification is highly demanded. Thus, off-line handwritten character recognition technology has Important practical value.The character recognition process contains pre-processing, feature extraction, identification, and so on major steps. Because of the structure of neural tree, It also has a feature extraction function. When tree structure is phylogeneticed ,some important characteristics will be copied and transimission as a leaves sub-node, eventually evolved to be the neural tree leaf node is an important feature to extract. So this paper use neural tree to identify for offline handwritten characters , to further improve the efficiency of character recognition.This study aimed to design recognition mode based on neural tree, the object for identify is a specific character set of English characters(26 uppercase letters of the alphabet). The main research work is organized around the handwritten character recognition technology based on neural tree. First of all, this paper summarizes the research backgroud and current situation of character recognition. Secondly analyzed the pre-processing of handwritten characters images including smooth,thinning and so on. Reseach the basic theory of neural tree and how to used it to recognition handwriting characters.the reseach focuses on tree structure optimization algorithm and its parameters optimzation algorithm. Finally set up a off-line handwritten recognition system. The system mainly composed by the pre-processing module, feature extraction module and the neural Tree identfication module.The main job of this pape include following aspects:(1)Set up a nerual tree identify model,and make a simulation testing in matlab environment. The effect of its identifcation were compared with the BP network's. According to test results,analysis of the perfomance of recognition model. Indication that the nerual tree mode improved the efficiency of handwritten character recognition. (2)Through simulation testing at vc++.net environment, set up a handwritten character recognition system based on neural tree. From character input,pre-processing,feature extraction,computer recognition to output the results,finished the whole process of recogniton.The experimental results show that in handwritten character recognition process can handle amounts of data quickly and achieve a good recongnition effect.