Multiple feature cooperation based handwritten Uighur character segmentation on mobile phone
Jing Li, Zhaoyang Lu, Adili Yimiti, Fuxiu Tan, Wei Wang · 2010
Robust and accurate handwritten character segmentation is the essential step for automatic Uighur word recognition on mobile phone. However, it's a hard work to do character segmentation due to Uighur's cursive style and other characteristics. In this paper, a hybrid multiple feature cooperation based algorithm is presented, the main characteristics of the proposed approach include: (1) A multi-feature analysis method is given to extract connectivity, ownership, location and local peak characteristics, and all the features are cooperated to guide the segmentation; (2) A joint character segmentation model is presented to achieve the final segmentation results. In order to evaluate the performance of the proposed method, we have developed a handwritten Uighur character capture system on mobile phone, in addition, a labeling system is also designed to get the ground truth of character segmentation. Experimental results with difficult handwritten Uighur characteristics demonstrate the effectiveness of the algorithm.