摘要
针对现有的线性判别分析算法中存在的降维舍弃空间的判别信息丢失问题,以及秩空间和零空间的判别信息难以兼顾的问题,本文提出了一种完全判别分析算法.算法通过新构建一个子空间以及其中的判别矩阵,实现了可以充分使用全空间的判别信息;且新空间的维数较低,算法流程简单,计算代价较小.相关实验结果证实了本文算法较传统判别分析算法有更好的性能和效率.
This paper proposes a complete discriminant analysis algorithm to solve the problem that existing linear discriminative analysis algorithms always miss discriminant information when reducing dimensions, and the problem that it is difficult to juggle discriminant information both in rank space and null space. This algorithm makes full use of all discriminant information in all space, by constructing a new subspaee and its diseriminant matrix. The dimension of the new subspaee is low and the algorithm is simple and costless. Theexperimental results demonstrate that this algorithm has better performance and efficiency than traditional discriminant analysis algorithms.
出处
《中国科学:信息科学》
CSCD
2012年第9期1181-1190,共10页
Scientia Sinica(Informationis)
基金
国家重点基础研究发展计划(批准号:2011CB302400)资助项目
关键词
判别分析
子空间
秩空间
零空间
人脸识别
discriminant analysis, subspace, rank space, null space, face recognition