A hybrid model for handwritten Chinese characters recognition

In this paper, a hybrid model is presented for handwritten Chinese characters recognition. First, a new and reliable stroke extraction method is proposed. Based on the extracted strokes, the cross-points can also be determined. The characters with the same numbers of strokes and cross-points are grouped together in rough classification stage. Second, a Bayesian neural network for fine classification is presented. The experimental results show that the commonly used 1000 characters can be partitioned into 70 groups and each group contains 23 characters on the average and the average recognition rates of 95% and 85% for recognizing the characters written by two specified persons and selected from the CCL/HCCR1 database had been obtained.<>