摘要
为进一步有效提升稀疏表示人脸识别系统的识别率和可靠性,在分析人脸图像稀疏表示系数分类能力的基础上,提出了一种基于残差加权的稀疏表示人脸识别新方法.该方法通过对类残差图像关于所属类稀疏表示系数的l2范数进行归一化加权,有效提升了原始基于类残差判决的识别能力.仿真实验结果表明:改进的基于残差加权的稀疏表示方法能够有效提高系统的识别性能.
To further enhance the performance of SRFR, an improved sparse representation classification (SRC) method for face recognition based on weighted residuals (WR) is proposed on the basis of analyzing the classification capability of sparse representation coefficients. The recognition capability of the original SRC is efficiently promoted by WR with 12- norm of sparse representation coefficients. Simulation experimental results show that the recognition performance of WR based SRC can he considerably increased.
出处
《中南民族大学学报(自然科学版)》
CAS
2012年第3期72-76,共5页
Journal of South-Central University for Nationalities:Natural Science Edition
基金
国家自然科学基金资助项目(60972081)
湖北省自然科学基金重点资助项目(2009CDA139)
中南民族大学中央专项基金(CZY12006)
关键词
人脸识别
稀疏表示
残差加权
face recognition
sparse representation
weighted residuals