摘要
为解决基于表示理论的分类法未考虑噪声样本对重构系数影响的不足,利用局部约束协同表示法改进最小二乘回归分类法,提出局部强化最小二乘回归分类法.该方法通过非负稀疏表示自适应选择近邻样本,并利用近邻样本的协同作用强化重构系数使得局部强化最小二乘回归分类法具有较好的鲁棒性和容噪性.该方法可以克服传统分类方法存在的过拟合问题.在4个人脸图像数据集上的实验结果表明该方法可以提高人脸识别准确率.
In order to improve the weakness of classification method based on the representation theorythat ignore the influence of noise on the reconstruction coefficients,we propose a local strengthen least-square regression classification method to improve the least-square regression classification method by using the local-constraint cooperative representation.The proposed method can select neighbor samples adaptively by using nonnegative sparse representation.It strengthens reconstruction coefficients by using neighbor data samples,and improves anti-noise ability.Furthermore,it can overcome the over-fitting problems that plague traditional classification methods.Experimental results on the four face recognition datasets show that this method can improve recognition accuracies.
作者
简彩仁
夏靖波
JIAN Cairen;XIA Jingbo(School of Information Science & Technology,Xiamen University Tan Kah kee College,Zhangzhou 363105,China)
出处
《厦门大学学报(自然科学版)》
CAS
CSCD
北大核心
2019年第1期122-126,共5页
Journal of Xiamen University:Natural Science
基金
福建省自然科学基金(2018J01101)
关键词
人脸识别
最小二乘回归
局部强化
自适应
分类
face recognition
least square regression
local strengthen
adaptive
classification