Abstract: To address the performance degradation of direction of arrival (DOA) estimation under snapshot-limited conditions in underwater acoustic environments, a sparse reconstruction method based on the multiple measurement vector (MMV) model is investigated. An extended MMV-orthogonal matching pursuit (OMP) algorithm is developed by jointly processing multi-snapshot observations and exploiting the joint sparsity of signals to achieve iterative support set refinement, thereby improving DOA estimation accuracy under low signal-to-noise ratio (SNR) conditions. Comparative simulations are conducted under varying SNR levels, and the performance is quantitatively evaluated using root mean square error (RMSE) and detection probability (DP). The results show that, with 10 snapshots and an SNR of −4 dB, the proposed algorithm improves the detection probability by approximately 2.2% and reduces the RMSE by about 1.01° compared with benchmark methods, while maintaining low computational complexity. Experimental validation based on anechoic water tank data demonstrates that the proposed algorithm produces continuous and stable DOA estimation trajectories with well-focused spectral peaks, enabling effective tracking of target direction variations. The results indicate that MMV-OMP achieves high estimation accuracy and strong noise robustness under snapshot-limited conditions, and exhibits promising potential for practical applications.