The measurement matrix plays an important role in compressed sensing theory. However, the measurement matrix is only dominated by the compressed signal, so it is influenced by the signal reconstruction significantly. Therefore, an optimization algorithm for the measurement matrix with the bat algorithm is introduced. In this algorithm, the measurement matrix is initialized by using the Gaussian matrix, and then updated with the bat algorithm, while the signal reconstruction error is taken as the objective function. To verity the performance, the proposed algorithm is compared with five other algorithms. Simulation results show that the proposed algorithm has better stabity and reconstruction accuracy.