摘要
稀疏表示近年来在模式识别领域已经取得了成功的应用,如目标探测和分类。稀疏保留投影(SPP)算法是基于稀疏表示理论所提出的一种特征提取方法,目标是获得一个线性投影空间,使得样本之间的全局重构关系得以在低维空间保留。SPP算法无需选择任何模型参数,具有很强的适应性,其灵活性及有效性在人脸识别中得到了详细的验证。文中结合二维Gabor小波与SPP算法用于人脸识别系统,二维Gabor小波主要用于提取人脸图像特征,SPP对图像特征进行降维。最后,在ORL人脸数据库上的实验表明,该算法较传统方法以及单独使用SPP算法的方法有较好的识别结果。
Recently,sparse representation has been successfully applied to solve many practical problems in the pattern recognition,such as target detection and classification. Sparsity Preserving Projections （ SPP） algorithm which is based on sparse representation theory has been put forward as a method of feature extraction,aimed at achieving a linear projection space,and making the global reconstruction rela-tionship between samples preserved in low-dimensional embedding subspace. SPP algorithm needs not to choose any model parameter, with strong adaptability,its flexibility and effectiveness is verified in the face recognition in detail. In this paper,combined 2D Gabor wavelet and SPP algorithm for face recognition system,2D Gabor wavelet is mainly used for face image feature extracting,SPP for di-mension reduction of image characteristics. Finally,in ORL public face database,experiments show that the algorithm has better recogni-tion result than traditional methods and the method of single SPP algorithm.
出处
《计算机技术与发展》
2014年第11期110-113,共4页
Computer Technology and Development
基金
江苏省自然科学基金(BK2011789)
东南大学毫米波国家重点实验室开放课题(K201318)
关键词
模式识别
特征提取
稀疏表示
稀疏保留投影
GABOR小波
降维
pattern recognition
feature extraction
sparsity representation
SPP
Gabor wavelet
dimension reduction
作者简介
陈静（1988-），女，硕士研究生，研究方向为图像处理与模式识别；
邱晓辉，教授，通信作者，研究方向为信号与信息处理与模式识别。
