摘要
提出一种新颖的零空间判别投射(NDPE)的子空间人脸识别方法。基于局部保持映射(LPP)和非参数判别分析方法,NDPF能够同时编码人脸数据流形的几何和判别结构,并且通过在零空间中求解特征值来克服小样本尺寸问题。为进一步提高人脸识别的准确率,提出融合双树复小波变换(DTCWT)与NDPE的方法。实验结果表明,所提人脸识别方法在ORL、Yale和AR人脸数据库上均取得了较高的识别率。
A novel face recognition method called null space discriminant projection embedding(NDPE) is proposed.Based on locality preserving projections(LPP) and nonparametric discriminant analysis methods,NDPE can encode both the geometrical and discriminant structure of the face data manifold,and address the small sample size problem by solving an eigenvalue problem in null space.Dual tree complex wavelet transformation(DTCWT) is integrated with NDPE to improve the recognition performance.Experiment results on ORL,...
出处
《光电子.激光》
EI
CAS
CSCD
北大核心
2009年第9期1208-1213,共6页
Journal of Optoelectronics·Laser
基金
国家“973”计划资助项目(2002cb312200)
国家自然科学基金资助项目(60575036)