基于子空间追踪的人脸识别

针对现有的基于稀疏表示的人脸识别方法没有更新优化选择的原子的问题，提出一种基于子空间追踪的人脸识别方法。在稀疏编码过程中的原子选择步骤中，引入回溯迭代优化思想和多原子选择方案，通过移除可信度较低的原子来更新优化候选支撑向量中选择的原子，使选择的原子与待识别人脸图像具有最相似的结构，从而在该原子上的稀疏编码系数具有较好的人脸重构能力。实验证明，与基于正交匹配追踪（OMP）算法和基于OMP-cholesky算法的人脸识别相比，该算法在ORL和Yale B人脸数据库上的算法复杂度较低且识别率均提高了约5%。

Face recognition based on subspace pursuit

Against the disadvantage of haven’t update selected atoms in existing face recognition method based on sparse representation, this paper proposes a face recognition based on subspace pursuit. This algorithm introduces back iterative optimization method and polyatomic options in the atomic choice in sparse coding, by removing the candidate atoms with low credibility to make sure that the chosen atoms have the most similar structure with the identifying face image, so the sparse coding coefficient can reconstruct faces well. The experimental results show that this algorithm has lower algorithm complexity and boosts about 5% recognition rate on ORL and Yale B face database compared with Orthogonal Matching Pursuit algorithm（OMP） and the OMP-cholesky algorithm.