Research on the method of handprinted Chinese character recognition

This paper presents new methods of stroke extraction directly from a handprinted Chinese character dot matrix, and Chinese character classification according to its structural relation based on stroke features, as well as the construction of a recognition dictionary in handprinted Chinese character recognition. These methods are characterized by a simple algorithm, powerful classification, fast recognition and easy dictionary maintenance. Based on these methods, a full-page handprinted Chinese characters recognition system has been made with better than a 93.1% recognition accuracy for all the commonly used 3755 Chinese characters (first class)