摘要： 分析稀疏表示的人脸识别方法的基本原理，针对采用基于正交匹配追踪(OMP)的稀疏表示算法时，所获得稀疏系数存在负值的问题，提出一种改进的正交匹配追踪算法。通过对稀疏系数的大小进行直接约束，减少负值稀疏系数的产生及算法迭代次数，并提高人脸识别速度。在ORL人脸数据库中的实验结果证明，改进后算法的识别率比原有算法提高了3%，迭代次数设置为7次最为合理。

关键词: 人脸识别, 正交匹配追踪, 稀疏表示, 稀疏系数, 信号重构, 压缩感知

Abstract: The fundamental principle of face recognition based on sparse representation is analyzed. When sparse representation algorithm based on Orthogonal Matching Pursuit(OMP) is adopted, the obtained sparse coefficient can be negative. To solve this problem, an improved orthogonal matching pursuit algorithm is presented. By directly limiting the sparse coefficient, the proposed method can reduce the amount of negative sparse coefficient and iterations. Meanwhile, face recognition is speeded up. Experiment on the ORL database show that the recognition rate of improved algorithm is 3% higher than the original one, and the most suitable iteration times is 7.

Key words: face recognition, Orthogonal Matching Pursuit(OMP), sparse representation, sparse coefficient, signal reconstruction, compressed sensing

中图分类号: 

• TP301.6