基于FDDL和混合稀疏的人脸识别
作者:王威 陆俊
来源:《科技视界》2017年第bd留置针26期
【摘 要】稀疏表示在人脸识别中取得了重大的突破。作为该领域的研究之一,FDDL字典学习寻一组不含冗余噪声等未确定因素的基,在对测试样本线性重构和分类方面都起到关键性作用;混合稀疏分类模型既实现了整体稀疏性,也考虑了不同类的字典原子之间的关系。将两者进行结合,在ORL库上的实验结果表明,此方法可显著提高识别率。 【关键词】稀疏表示;人脸识别;FDDL字典学习高新技术产业化;混合稀疏
office2002中图分类号: TP391.41 统一身份认证服务文献标识码: A 文章编号: 2095-2457(2017)26-0087-002厦门集美大桥
Face Recognition Based on FDDL and Hybrid Sparse
WANG Wei1 LU Jun2
(1.School of Electrical and Information Engineering, Anhui University of Science an大陆漂移说
d Technology, Huainan, Anhui 232001, China;2.State Grid Anhui Electric Power Company Information Communication Branch, Hefei 230061, Anhui, China)
【Abstract】Sparse representation has made a major breakthrough in face recognition. As one of the research fields in this field, FDDL dictionary learning to find a group of unconfirmed factors such as redundant noise plays a key role in the linear reconstruction and classification of test samples. The hybrid sparse classification model not only achieves The overall sparsity also takes into account the relation between different classes of dictionary atoms. The experimental results on the ORL database show that this method can significantly improve the recognition rate.