Handwritten Character Recognition Algorithm Based On Template Matching Method

The development of human civilization can’t be separated from the transmission of messages. Text message has been an important medium for people to transmit information which is charged with the important responsibility of recording human civilization. With the development of science and technology, the accumulation of text message is increasing. How to efficiently store and make use of these text messages becomes is necessary question to be handled. Optical character recognition provides a solution to this problem. As an important branch of optical character recognition, handwritten numeral recognition has been widely used in the fields of finance, postal, medical, transportation, educational and has got more and more attention. At present, there are many handwritten numeral recognition method, but very few can be actually used practice life for handwritten numeral recognition requires almost100%recognition rate.On the basis of the theory of image pattern recognition, an enhanced template matching algorithm based on character’s geometric outline feature is proposed in this paper to meet the requirements of high correct recognition rate and recognition accuracy. An information input system which recognizes handwritten characters of hospital information tables and campus hospital prescriptions is also created. The main work of this paper is as follows:(1) The global geometric outline feature is used as a crucial basis for classifying a character. A simple and easy operating enhanced template matching method is exploited to design the classifier. The introduction of scaling factor and coefficient of variation increase the classification accuracy and reliability. To accelerate the processing speed, the design of the template base is improved. A much smaller template base using standard font numerals instead of handwritten numerals is created.(2) An information input system is created by use of VC++6.0programming software. In this system, special character location algorithm based on image matching and multiple characters segmentation algorithm based on checkbox or blank pixel column are designed direct at handwritten numerals in hospital information tables and campus hospital prescriptions. This information input system can automatically recognize characters, display and store the data which significantly improves the efficiency.(3) Experiment results indicate that algorithm effectively increases the correct recognition rate, accelerates the speed of recognition and renders good robustness for the size and location of the characters in the image. The implementation of the information input system validates the feasibility and effectiveness of the algorithm.