Abstract:

In order to solve the problem of automatic segmentation of handwritten Chinese text , a dynamic program-ming-based online handwritten Chinese character segmentation method is proposed. In this method , the geometrical features of handwritten strokes , the stroke sequence information and the class probability given by the neural net-work classifier are used to construct a segmentation cost function , and are then applied to the pre-segmentation and recognition-based segmentation stages of handwritten sentences , respectively. After that , the dynamic programming algorithm is adopted to find the optimal segmentation path. The pre-segmentation can effectively reduce the amount of segmentation hypothesis with a reasonable incorrect segmentation rate , thus speeding up the segmentation. Ex-perimental results show that the pre-segmentation stage achieves an incorrect segmentation rate of O. 579% and an over-segmentation rate of 11. 1 % , and that the final correct segmentation rate is 88. 2 % without using any language model.

Key words: character recognition, dynarnic programming, online handwritten Chinese character segmentation, geo-metrical feature