The Research of Algorithm for Handwritten Character Recognition in Correcting Assignment System

Handwritten character recognition is the key technique in correcting assignment system as well as development of aided instruction software. Considering the disparity in distribution of the pixel, we propose a distribution-based algorithm for handwritten character recognition. Based on the theory of Image Segmentation, the centroid of a character can be found. Around this centroid, the image is divided into equal angle regions clockwise and the direction feature of the character distribution can be obtained. The hand-writing restraint will be flexible since the method named Deflection Correction is adopted, as well as the matching error will be reduced. In the principle of high-accuracy matching, we use the minimal matching database to approach the real-time character match. The algorithm provides us a satisfactory recognition rate, especially on numbers and English characters, and the results of recognition can be classified effectively which are defined as Learning Functions. Without necessity of thinning and finding the starting location and direction of the character, this method directly computes the regularity of distribution in each region of a character image. It overcomes the complex process of extracting skeleton which traditional method has to adopt. With the advantage of speediness, accuracy and robustness, the system has an excellent vista.