Abstract: In order to improve the classification accuracy of sparse representation classification algorithm for samples belonging to the same direction but different categories, a sparse representation classification algorithm based on Volterra kernel optimization is proposed. Firstly, the original face images are divided into nonoverlapping blocks and mapped into high-dimensional space by Volterra kernel. In the training phase, Fisher criterion is followed, and the objective function is defined according to maximizing the distance between classes and minimizing the distance within classes to obtain the optimized Volterra kernel. The experimental results on ORL and YaleB standard datasets show that the classification accuracy of the SRC face recognition classification method based on Volterra kernel optimization is improved by 3%.