局部轨迹信息的联机手写维吾尔文单词过切分

On-line handwriting Uyghur words over-segmentation of local trajectory information

将手写单词分割成字符是一项艰巨的任务。本文提出一种基于局部轨迹信息的单词过切分方法。该方法中,在不预先对附加笔划进行分组,不确定基线位置和不做倾斜校正的情况下探索单词轨迹的局部信息进行过切分。利用的局部信息包括手写轨迹中的平直点、局部最大和最小、局部最右和最左。探索手写单词原轨迹点序列中的局部最大值和局部最小值的点,从中找出切分点。初步确定的切分块里找到最右边和最左边的点更新切分点。本文提出的方法在5个不同作者的500个无约束手写单词样本上进行测试。平均字符切分召回率和正确率分别为93.35%和69.97%。

Segmenting the handwritten word into characters is a tough mission for alphabetic scripts. This paper proposes an effective character over - segmentation method for online handwritten Uyghur words. This method tries to find segmentation points without pre - grouping of delay - strokes, determining baseline position and skew correction. The informative points including horizontal fiat, local pcak and valley, local rightmost and lefimost points for making pre -step over-segments. The local ma：dma and minima points are explored to find the segment points from the handwritten trajectory. The pre - determined segmentation points are modified by the rigbtmost and leftmost points within the pre - segmented fractions. The proposed method is tested on 500 unconstrained handwritten word samples from 5 diffm^nt writers. The average character segmentation recall rate and correct are observed as about 93.35% and 69.97%.