基于人脸分类和K_L变换的人脸识别新方法.pdfVIP

博士论坛 《微计算机信息》( 测控自动化)2010 年第26 卷第 3-1 期 文章编号:1008-0570(2010)03-1-0008-03 基于人脸分类和K-L 变换的人脸识别新方法 A new method of face recognition based on face classification and K-L transformation (东北大学) 高晓兴 李仁睦 王文佳 常桂然 GAO Xiao-xing LI Ren-mu WANG Wen-jia CHANG Gui-ran 摘要 为解决目前人脸识别领域中 对含有大量人脸样本训练库识别准确性低 实时效果差的问题 本文提出了一种基于人 : , 、 , 脸分类的 变换人脸识别新方法 该方法首先采用协同模式将 人脸库中的人脸样本分成若干聚类 之后利用 K-L 。 ORL ; K-L 变换的方式识别待测人脸样本 降低了参加识别比对的样本数量 提高了人脸识别的效率 适用于身份鉴别领域 。 , , 。 关键词 人脸分类 人脸识别 特征提取 变换 : ; ; ; K-L 中图分类号: TP391.4 文献标识码: A Abstract: In order to solve the problem of low accuracy and poor real-time effect in training database consisted of mass of face sam- ples, we have proposed a new K -L transformation method for face recognition which is based on face classification. For one thing, this method use synergistic model to divide the face samples in ORL face database into a number of clustering; then, use K-L trans- 技formation to recognize the test samples. This method reduced the number of face samples to be recognized and cut down the number of samples to be identified while increased the efficiency of face recognition, this method can be applied to identity verification. 术Key words: Fa