The invention relates to a direction-of-arrival (DOA) estimation method based on off-grid sparse Bayesian, which includes the following steps: S1, initializing a sparse base matrix: Phi(beta)=A+Bdiag(beta); S2, updating the mean and variance of estimation signals based on the sparse base matrix, and updating and iterating the noise variance, signal variance and off-grid parameters based on the mean and variance of estimation signals; S3, judging whether the maximum value of the absolute values of the iterated and updated off-grid parameters is greater than a set threshold C, executing S4 if the maximum value is greater than the set threshold C, otherwise, executing S5; S4, migrating the grid according to the off-grid parameters and the threshold C, updating matrixes A and B, and updating the sparse base matrix; S5, judging whether convergence is achieved or the number of iterations reaches the upper limit, returning to S2 if convergence is not achieved and the number of iterations doesnot reach the upper limit, otherwise, ending iteration; and S6, using the updated sparse base matrix to update an estimation model of off-grid DOA to get an estimation result. Compared with the priorart, the method of the invention has the advantage of high accuracy.