摘要： 延迟笔画处理是维吾尔文联机手写体单词识别中的难点，样本数据格式为一串具有时序性的序列，造成字母主笔画和延迟笔画在序列位置上多半不连续，很大程度上增加了延迟笔画处理和单词识别的难度。为此，提出一种维吾尔文延迟笔画处理方法，其中包括延迟笔画查找与投影，实验结果证明该方法具有良好的效果，单词识别准确率达到93.71%。

关键词: 维吾尔文, 延迟笔画, 手写体识别, 联机识别, 隐马尔可夫模型

Abstract: Processing delay stroke is a difficulty in online Uyghur handwriting recognition. Because sample data format is a timing sequence, it makes letter’s main stroke and delay strokes uncontinuous in sequence. It greatly increases the difficulty of processing delay stokes and recognizing handwriting word. To solve this uncontinuous problem, this paper proposes a delay stroke processing method which contains seeking and projection of delay strokes. Result shows this method is robust and recognition accuracy rate is up to 93.25%.

Key words: Uyghur, delay stroke, handwriting recognition, online recognition, Hidden Markov Model(HMM)

中图分类号: 

• TP391