Abstract: Kernel Nearest Neighbor Convex Hull（KNNCH） classifier involves solving convex quadratic programming problems,which requires large memory and long computation time for large-scale problem.Therefore,it is important for KNNCH classifier to reduce the computation complexity without degrading the prediction accuracy.This paper present a named Kernel Subspace Sample Selection（KSSS） method to choose training samples for KNNCH classifier.KSSS algorithm is an iterative algorithm in one class,which selects the furthest sample to the subspace of the chosen set at each step in kernel space.The experiments on the training-synthetic subset of the MIT-CBCL face recognition database show that our KSSS+KNNCH approach could reach 100% recognition rate with less samples and much faster test speed than KNNCH.