摘要
针对目前主流人脸识别算法,在单样本条件下,当性能严重下降根本无法工作时,提出了一种结合Gabor和核监督局部保持投影的单样本人脸识别算法。选取数据库中中性表情的近正面人脸图像作为训练样本,通过几何变换产生15幅虚拟样本,对每幅样本图像提取Gabor特征,采用核监督局部保持投影方法进行特征提取,欧氏距离最近邻分类器进行分类。根据ORL数据库、Yale数据库和FERET数据库上的实验结果表明,核监督局部保持投影(GKSLPP)算法具有较好的识别效果。
Many reported face recognition techniques rely heavily on the size and representative of training set, and most of them will suffer serious performance drop or even fail to work if only one training sample per person is available to the systems. This paper proposed a new algorithm based on Gabor and kernel supervised locality preserving projection. In this paper, we firstly choose one neutral-expression and nearly frontal face im- age from images per person in the databases as the training sample, and then we generate 15 virtual images of the selected images by geometric transformations. Secondly, we use Gabor and kernel supervised locality pre- serving projection method for feature extraction and Euclidean the nearest neighbor classifier for classification. Results on FERET database, ORL database and Yale database verify the effectiveness of this algorithm in the face recognition using a single training sample.
出处
《信息化研究》
2012年第1期49-52,共4页
INFORMATIZATION RESEARCH
关键词
人脸识别
单样本Gabor小波变换
核监督局部保持投影
face recognition
single sample problem
Gabor wavelets
kernel supervised locality preser-ving projection
