The invention discloses a sparse adaptive DOA estimation method and system based on a pseudo-inverse perception dictionary, and the method comprises the steps: receiving K narrow-band signal sources through an M-element uniform linear array, and obtaining an observation vector of a noise-free field; taking the observation vector as an initial residual error, adopting a sparsity adaptive orthogonalmatching pursuit method based on a pseudo-inverse perception dictionary, obtaining a residual error value corresponding to the current stage by continuously iterating the step length, obtaining a minimum residual error value according to the convergence comparison of the residual error values of different stages, and obtaining a corresponding sparse coefficient; and obtaining the directions of the non-zero elements in the K narrowband signal sources. According to the method, the convergence value of each step iteration is compared, the step corresponding to the minimum convergence value is used as the sparsity, the threshold error is effectively eliminated, and the application range is expanded; the advantages of the perception dictionary are combined, and a new design criterion of the pseudo-inverse perception dictionary is provided, so the error influence caused by atom strong correlation under the condition of high resolution is overcome, and the direction of the incoming wave is positioned more accurately.