一种基于局部和判别特性的降维算法  

Dimension reduction algorithm based on locality and discriminant characteristics

作　　者：张国印[1] 楼宋江 王庆军[1] 程慧杰 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,哈尔滨150001

出　　处：《计算机应用研究》2009年第9期3324-3325,3329,共3页Application Research of Computers

摘　　要：提出了一种基于LPP和LDA的降维算法。该算法不仅考虑了LPP能保持局部邻近关系属性,还考虑了LDA能使降维后的数据更易于分类属性,并且该算法是线性的,很容易将新样本映射到目标空间。在人脸识别中的实验验证了算法的正确性和有效性。This paper proposed a dimension reduction algorithm based on LPP and LDA, which not only utilized that LPP could preserve locality, but also took into account that the LDA made the data more discriminant and thus made classification mare easier. The proposed algorithm was linear, which could map the out of sample data to the target space easily. Experiments in face recognition validate the correctness and effectiveness of the algorithm.

关 键 词：维数约简 局部保持投影 线性判别分析 人脸识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]