The invention discloses an acoustic vector array DOA estimation method, which comprises the steps that: signals sent by L remote underwater acoustic targets are narrowband signals at frequency f, a receiving signal array is a uniform linear array comprising M vector sensor array elements, and an array element distance is half wavelength of the transmitted signals; an entire underwater acoustic target space is divided into K spatial positions, and each spatial position corresponds to one directional angle; a sparse bayesian learning DOA estimation algorithm is adopted for the signal Sw<~>, a maximum posterior probability of a signal source is obtained to achieve azimuth angle estimation of the targets through solving a value of a hyper-parameter, the hyper-parameter is subjected to iterative calculation till convergence, and an underwater acoustic target sparse reconstructed signal vector is calculated to be Si<~> through the final iterative calculation; a position of a S<~> non-zero row is determined, a non-zero element position of the sparse reconstructed vector S<~> corresponds to an actual DOA angle, and the DOA estimation is completed finally. The acoustic vector array DOA estimation method can improve DOA estimation accuracy, obtains more incisive directional beams and lower side lobes, and achieves omnibearing DOA estimation.