摘要
为进一步提高基于稀疏表示的人脸识别方法识别率,提出一种基于样本扩充的核稀疏表示方法 (KSRMSE)。通过在原始样本中添加少量的噪声,扩大原始样本集的规模,使用核诱导函数从训练样本集中挑选N个最近邻样本,利用这N个最近邻样本的线性组合表示测试样本,根据表示的结果对测试样本进行分类,通过修改N值获得更高的分类精度。实验结果表明,相比同类识别算法,该方法具有更好的识别效果。
To improve the recognition rate of face recognition method based on sparse representation,a kernal sparse representation method based on samples expansion method（KSRMSE）was proposed.The training samples were extended to form a new training set by adding some noise to them and a kernel-induced function was used to determine Nnearest neighbors of the testing sample from the total training samples.The testing sample was represented using the linear combination of determined Nnearest neighbors and the classification was implemented according to the representation results.Through the different values of Nset,the classification was more accurate.Experimental results show that KSRMSE can get better classification results than the same type of algorithms.
出处
《计算机工程与设计》
北大核心
2016年第5期1357-1361,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61170126
61340037)
江苏省普通高校研究生科研创新计划基金项目(CXLX13_67)
南通市科技计划应用研究基金项目(BK2012038)
关键词
人脸识别
稀疏表示
核诱导
样本扩充
N最近邻
face recognition
sparse representation
kernel-induced
samples extension
N nearest neighbors
