基于多投影的脱机手写藏文字符特征提取方法

Feature Extraction Method of Off-line Handwritten Tibetan Character Based on Multiple Projection

脱机手写藏文字符的识别能够促进藏文化的发展和传播,其识别的方法是根据脱机手写藏文字符图像的特征进行识别。由于原始定义下的行列投影向量对于脱机手写藏文字符图像的表示不够充分,文中提出一种基于多投影归一化的脱机手写藏文字符特征提取算法。首先,对脱机手写藏文字符图像分别在横向、纵向、主对角线方向和次对角线方向进行投影,得到行投影向量、列投影向量、主投影向量和次投影向量;然后,对投影向量进行归一化处理,将归一后的向量合并成一个向量,即为该脱机手写藏文字符图像的特征向量;最后,使用KNN分类器对测试样本进行识别。对建立的脱机手写藏文字符样本数据库中的样本进行实验。结果表明,该算法不仅计算简单,而且有较好的识别效果。

Off-line handwritten Tibetan character recognition can promote the development and propagation of Tibetan culture, and the method of classification is based on the feature of off-line handwritten Tibetan character image. Since original definition of projection entropy does not make full use of image information, a feature extraction method of off-line handwritten Tibetan character based on multiple projection normalization is proposed. Firstly, an off-line handwritten Tibetan character image is scanned in horizontal, vertical, main diagonal and secondary diagonal directions to create a row projection vector,a column projection vector,a main projection vector and a secondary projection vector. Secondly, all projection vectors are normalized to create a multiple projection normalized vector that is the feature vector for this character image. Finally, KNN classifier is used in classification. The proposed method is tested on the developed off-line Tibetan handwritten character sample database. The experimental results demonstrate that the proposed method is not only easy in calculation but also efficient in recognition accuracy.