摘要： 提出了一种完整的多字体印刷藏文字符识别方法:先提取输入字符的方向线素特征组成特征向量,然后采用两级分类策略判定字符类别。该方法在训练集和测试集上的识别率分别达到99.73%和99.44%,证明了其有效性。

关键词: 藏文字符识别, 方向线素特征, 带偏差欧氏距离

Abstract: A comprehensive algorithm for multi-font printed Tibetan character recognition is proposed in this paper. Directional line element features are extracted from the normalized input character to formulate feature vector. Then two-stage classification strategy is adopted to determine the category of input character. The recognition rates on training set and test set reach 99.73% and 99.44%, respectively. The experimental results show the validity of proposed algorithm.

中图分类号: 

• TP391.4