OFFLINE RECOGNITION OF CHINESE HANDWRITING BY MULTIFEATURE AND MULTILEVEL CLASSIFICATION

One of the most challenging topics is the recognition of Chinese handw riting, especially off line recognition. In this paper, an oft line re cognition system based on multifeature and multilevel classification i s presented for handwritten Chinese characters. Ten classes of multife atures, such as peripheral shape features, stroke density features, an d stroke direction features, are used in this system. The multilevel c lassification scheme consists of a group classifier and a five-level c haracter classifier, where two new technologies, overlap clustering an d Gaussian distribution selector, are developed. Experiments have been conducted to recognize 5,401 daily-used Chinese characters. The recog nition rate is about 90 percent for a unique candidate. and 98 percent for multichoice with 10 candidates.