Four Corner Code based pre-classification scheme for Chinese character recognition

摘要

One of the important step in handwritten Chinese character recognition is the pre-classification. In this paper, a simple pre-classification scheme based on four corner coding is proposed. Four corner code utilizes the information abundant area in Chinese character which is located on the peripheral. Multiple four corner codes are generated to cater for writing variations. Moreover, in case of mis-classification, a feedback model is established to increase the pre-classification rate. The idea of the feedback model is to identify the corner code stroke which is the highest probable to be mis-identified. The recognition system then re-generates another set of four corner codes. Experimental results revealed that a 92% pre-classification rate can be attained.