Learning Pattern Generation for Handwritten Chinese Character using Pattern Transform Method with Cosine Function

In pattern recognition, the number and quality of learning patterns is of crucial importance. When the number and quality of learning patterns are limited, error occurs in the presumed distribution of patterns and the precision of whole recognition system decreases. In this paper, a new pattern generation method is proposed which contributes to improvement of the performance of a handwritten Chinese character recognition system. By using this pattern generation technique, we increase the number and quality of learning patterns by using transform method with cosine function. Patterns generated this way are then selected using pattern selection method and the patterns unsuitable for learning are discarded. The recognition experiment on HCL2000, a handwritten Chinese character database, shows that our method improves the recognition precision of whole system