基于实值Gabor能量特征的手写体维文字符识别  

Recognition of handwritten Uyghur character based on real value Gabor energy feature

作　　者：姜文[1] 刘立康[1] 

机构地区：[1]西安电子科技大学通信工程学院,陕西西安710071

出　　处：《计算机工程与设计》2013年第9期3273-3278,共6页Computer Engineering and Design

摘　　要：提出了一种基于实值Gabor滤波器手写体维吾尔文字符特征提取算法。将手写体维吾尔文字符图像进行滤波处理之后,在将图像进行分块,提取出每一块的实值Gabor能量值。由这些能量值形成一个能量矩阵,将矩阵降维之后得到字符的特征相量。完成特征提取后,使用KNN识别分类器进行识别。对手写体维吾尔文单字符数据库中的样本分别进行基于实值Gabor能量特征的手写体维吾尔文字符特征识别和字符笔迹特征识别。对KNN分类器识别的平均识别率和平均候选识别率进行了数据分析。实验结果表明,该算法简单有效且识别率比较高。An algorithm of energy features of hand-written Uyghur character based on real value Gabor filter is presented. After filtered the image of hand-written Uyghur character, the image is divided into small pieces, the energy of each pieces is extrac- ted. These energy values can form a energy matrix and reduce dimension as the characteristic matrix after phasor. After feature extraction, KNN classifier is used in character recognition. The samples are recognized in the database of the hand-written, Uyghur character using the method based on real value Gabor energy feature in both feature recognition and writer feature recog- nitition. The experimental result shows, the algorithm is simple, effective and have a remarkable recognition rate.

关 键 词：手写体维吾尔文字符 GABOR滤波器 KNN识别分类器 训练样本 测试样本 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]