A STUDY ON MULTI-FONT PRINTED CHINESE CHARACTER RECOGNITION SYSTEM
Hu W · Zhongwen xinxi xuebao · 1988
An approach to multi-font printed Chinese character recognition is proposed in this paper. The problems of inputting image of characters, preprocessing, character segmentation, feature extraction as well as character classification have been discussed. Accordiog to the characteristics of multi-font printed Chinese characters, the number of cutting across strokes, the external and internal areas within a character are used as features for multi-stage classifcation. Experiments show that this method has the advantages of overcoming noises, displacements of characters and it can be achieved identical recognition result when a character appears in different fonts, on this basis, a multi-font printed Chinese character recognition system has been built, which can recognize 3755 most frequently used printed Chinese characters with SONG and BLOD FACE fonts mixed in one document, with a same dictionary .Its average recognition rate is more than 99%.An approach to multi-font printed Chinese character recognition is proposed in this paper. The problems of inputting image of characters, preprocessing, character segmentation, feature extraction as well as character classification have been discussed. Accordiog to the characteristics of multi-font printed Chinese characters, the number of cutting across strokes, the external and internal areas within a character are used as features for multi-stage classifcation. Experiments show that this method has the advantages of overcoming noises, displacements of characters and it can be achieved identical recognition result when a character appears in different fonts, on this basis, a multi-font printed Chinese character recognition system has been built, which can recognize 3755 most frequently used printed Chinese characters with SONG and BLOD FACE fonts mixed in one document, with a same dictionary .Its average recognition rate is more than 99%.