Abstract: Because traditional subspace algorithms have low direction of arrival (DOA) estimation accuracy or cannot be estimated under the conditions of few snapshots, low signal-to-noise ratio, and source coherence, the application of compressed sensing theory in DOA estimation was studied in this paper to solve the problem that sparsity adaptive matching pursuit algorithm cannot be effectively estimated in a noisy environment and the selection of a large initial step causes overestimation. On the basis of this algorithm, the iteration termination conditions of the algorithm were improved and optimized by using the variation rule of the iteration residuals. The adaptive adjustment of the step size can quickly and accurately approach the source sparsity. The simulation results show that the proposed algorithm has the advantages of high estimation accuracy, fast running speed, and good robustness to noise. It promotes the further fusion of compressed sensing and DOA estimation under actual conditions.