Abstract: The typical Sparse Representation for Classification（SRC）based on [L1] norm minimization is a very popular pattern recognition method due to its desirable classification performance in many applications. Nevertheless, the traditional SRC method usually suffers from the low computational efficiency. In order to deal well with this problem, this paper proposes an efficient and effective approach based on coordinate descent algorithm, which can significantly enhance the computational efficiency and achieve good classification results. The experiments on popular face databases demonstrate that the proposed approach is promising.

Key words: sparse representation, coordinate descent method, classification, face recognition

摘要： 经典的稀疏表示分类（Sparse Representation for Classification，SRC）算法是一种基于[L1]范数最小化问题，它在很多应用场合都能取得很好的分类效果，是目前备受关注的一类识别算法。然而，传统的SRC算法在求解[L1]范数最小化问题时，往往计算效率比较低。为有效解决这个问题，提出了一种快速有效的分类算法，它利用坐标下降方法来实现SRC算法。该方法既可以显著地提高计算效率，又可取得较好的分类结果。在不同人脸库上的实验表明，所提的算法具有良好的应用前景。

关键词: 稀疏表示, 坐标下降算法, 分类, 人脸识别