摘要
稀疏保留投影是一种有效的特征提取算法,它虽然能很好地保留样本间的稀疏重构特性,但是得到的特征分量通常具有统计相关性,存在数据冗余。为此,引入不相关约束条件,提出了不相关稀疏保留投影特征提取方法,利用推导出的公式提取不相关判别特征集,进而提高了识别率。在PIE、Extended Yale B和AR人脸库的实验结果表明:该方法有效且稳定,与MLHOSDA、SPP和LPP相比具有更高的正确识别率。
Sparse Preserving Projection is an effective feature extraction algorithm. Although it keeps sparse refactoring features between the samples well, the discriminant feature based SPP is generally statistical correlated, which makes the discriminant feature redundant. This paper added an uncorrelated constraint, and the projection feature extraction based uncorrelation sparse preserving projection method was proposed and theoretically derives the formula of extracting the uncorrelated discriminant features, so the recognition rate was improved. Ultimately experiments on PIE, Extended Yale B and AR face database show that the method in this paper is effective and stable, and has a higher correct recognition rate compared with the MLHOSDA, SPP and LPP.
出处
《重庆理工大学学报（自然科学）》
CAS
2016年第7期129-134,共6页
Journal of Chongqing University of Technology：Natural Science
基金
商洛学院科研基金资助项目(14SKY008)
关键词
特征提取
不相关
稀疏保留投影
人脸识别
feature extraction
uncorrelation
sparse preserving projection
face recognition
