Kernel sparse representation-based classification (KSRC) is currently one of hot research topics in the fields of pattern recognition and computer vision KSRC has excellent cla ssification performance,but is not able to obtain the important information about data locality,resulting in the f act that the discriminating power of sparse representation coefficients yielded by KSRC is reduced.Considering the i mportance of data locality,in this paper a new classification algorithm based on locality-sensitive kernel sparse representation is proposed for face recognition.The proposed method integrates both sparsity and data locality in t he kernel feature space so that it can obtain good discriminating sparse representation coefficients for classifica tion.Experimental results on two benchmarking face databases,i.e.,the ORL database and the Extended Yale B data base,demonstrate that the proposed method outperforms kernel sparse representation-based classification ( KSRC),sparse representation-based classification (SRC),locality-constrained linear coding (LLC),support vector machines (SVM),the nearest neighbor (NN),and the nearest subspace (NS).Therefore,the proposed method is able to achieve promising performance for face recognition.