Research And Implementation Of An Online Handwritten Chinese Character Recognition System

Unlike English,Chinese characters,which belong to two dimensions square characters,cannot be put into the computer directly.It is necessary for people to master codes and input by operating keyboard,which was once an only pattern for Chinese charcter inputing,at present,keyboard inputing is still a main method to input Chinese characters. But studying how to input by operating keyboard is not an easy thing,it requires to remember a number of codes,which makes it not convenient to many people. With the development of computer technology,a new input method named on-line handwritten Chinese character recognition appears.It is a real-time recognition method,by this method,computer can recognize Chinese characters while users are writing on a handwritten board.This method need not remembering codes and orperating keyboard.It's convenient and easy to learn.This dissertation mainly discusses the researches and implementations of an on-line handewritten Chinese character recognition system based on a mouse device,which can recognize Chinese characters recuring to mouse device.The system has little influence on cpu speed.It is easy to use.The methods of on-line handwritten Chinese characters recognition can be classified into two categories.One is based on Chinese character recognition,the other is based on stroke recognition.The dissertation adopts the latter,i.e.the method that recognize stroke firstly and recognize Chinese character secondly.Because Chinese character is made of stroke,and the characteristic of on-line recognition is recognizing each stroke while writing each stroke.Therefore,in the recognition system,stroke recognition is the preconduction and key of Chinese character recognition.The dissertation discusses the categories and characteristic of stroke firstly,then makes all kinds of statistics on Chinese character stroke information.The data is valuable and plays a significant role in the on-line handwritten Chinese character recognition system this dissertation researches.The stroke recognition method in this dissertation is on the basis of the conception— direction code.In this dissertation,a series of coordinates can be gained when we are moving the mouse.Because these coordinates are discrete,the curve between each twoadjoining data dot is the least beeline that the computer can differentiate.In this dissertation,there are 8 kinds of direction codes are defined,each segment is marked with a code.In this way,each stoke track can produce a sequence of direction code.Via filtering and mergering the sequence,we can obtain the stroke extraction sequence of this stroke,then adopt the dynamic programming method to recognize this stroke according to stroke recognition dictionary.Thereinto,partial stroke connection recognition is also implemented.Because of the randomicity of handwritten characters,handwritten stroke or Chinese characters are depended on user,time,writing condition,equipment and circumstance.In the course of recognition,the direction code sequence is stochastic and uncertain.So it is necessary for the system to possess self-learning ability,that is to say,when a new pattern appears,the system can learn and remember it.There are many classes of machine learing methods,the dissertation mainly discusses syntax learning and neural network method.Implementing stroke recognition is the first step of two-layer recognition method this dissertation adopted.The second step is implementing Chinese character recognintion.In the dissertation,we design a flow chart of the recognition system.There are mainly four problems we have discussed during Chinese characters recognition.The first one is the post processing problem for stroke recognition,two methods are proposed in this dissertation.The second one is,analyzing the recognition dictionary and relevant problems.The third one is the problem about matching stroke sequence.The last one is,discussing words association function of the recognition system.The spending of recognition system this dissertation researched is little. It can save software and hardware resource effectively.The system has implication in embedded systems and microcontrollers.It pocesses huge implication potential.