Abstract: To improve the extraction performance of weak ultrasound echo signals in low SNR environment, an optimized matching pursuit (MP) sparse decomposition algorithm is proposed. An adaptive particle swarm optimization (PSO) algorithm with continuous space search ability is introduced into the MP sparse decomposition to alleviate the contradiction between the ergodic limitation requirement and over-completeness of MP atomic set. By improving the parameter-adaptive setting of the PSO algorithm and the objective function and reconstruction function of the MP algorithm, the adaptive MP sparse decomposition algorithm improved by PSO optimization is realized. And then, a continuous over complete Gabor atom set is established, which improves the matching degree between the optimal atom and the sound signal in the evolution process. Finally, the echo signal is reconstructed by the optimal atom through the reconstruction function to realize noise reduction and echo accurate extraction. The experimental results show that the proposed algorithm significantly reduces the amount of computation, which is better than that of the existing wavelet thresholds method and others, and achieves better robustness.