K Nearest Neighbor Based Local Sparse Representation Classifier

The sparse representation-based classifier (SRC) has been developed and shows great potential for pattern classification. SRC is a global representation based method in that a testing sample is represented by all training samples. Thus, it is time-consuming when the training sample size becomes large. This paper presents a local SRC method, called KNN-SRC, which chooses K nearest neighbors of a testing sample from all training samples to represent the testing sample. Since K is much smaller compared to the total number of training samples, KNN-SRC is much faster than the global SRC. Moreover, using the K nearest neighbors to represent a testing sample can avoid the case that the testing sample is sparsely represented by the training samples that are actually parts of the testing sample but far away from it, so KNN-SRC could improve the performance of SRC. The proposed KNN-SRC is tested using the CENPARMI, NUST603, Yale B and AR databases. The experimental results show KNN-SRC is more effective and efficient than SRC.