Abstract: A method based on sparse representation for gait recognition was proposed using the CASIA_B and CUSD database. The gait silhouette was centralized and normalized first, then the AEI (active energy image) was calculated based on the previous operation and used as the feature image for gait recognition. Two methods were used to establish the dictionary and calculate the decomposition coefficients for the sparse representation: one for reconstruction, and the other for discrimination. OMP (orthogonal matching pursuit) algorithm was used for coefficients decomposition. The result of experiment shows that the proposed method can effectively recognize the gait, and the accuracy of the recognition is high and recognizing speed is much faster.