Abstract: A pattern recognition model with reject option, which is based on sparse representation combined with manifold distance hyperspherical covering model, is constructed in this paper. The samples in each class set can be supposed to distribute on a nonlinear manifold, so the local linear manifold subspace hyperspherical covering model for each class is obtained as the first step. And the input test pattern could be rejected or accepted by all the subspace boundaries associated with each class. Then, if a pattern is accepted by the above step, the sparse representation classifier (SRC) is used for classification in the training sample set. Experiments on the UCI database, the MNIST database of handwritten digitals, the MITCBCL face recognition database, and the CMU AMP face expression database show that this method is valid and efficient.