通过分析维吾尔文字母自身的书写特点,为了进一步提高识别率,在原先的外围轮廓特征基础上,再提取了附加笔划形状特征、字母上下左右密度比例特征、笔画数特征以及主体部分和附加部分的位置关系特征,进行了诸多特征有效组合下的字母识别实验。其中,最高识别率达到了75.7%,比原先在单个特征下的识别率提高了近10个百分点。这表明,不同特征的有效组合是可行的,而且具有继续研究的潜力。
The writing feature of Uyghur character is analyzed, and based on original peripheral contour feature, more features are extracted, including the shape feature of additional strokes, feature of bottom-up and left-right density ratio, stroke number feature, and location feature of additional part for Uyghur characters, thus to further improve the recognition rate. Meanwhile, several recognition experiments on the above features alone and their various combinations are also done. The highest recognition rate reaches 75.7%, and as compared with that under the original individual feature, the recognition rate is increased by lO%.This shows that the effective combination of different features is feasible, and has the potential for further study.