Abstract: To solve the problem of steering vector error in the adaptive beamformer, this paper presents a spectral analysis based robust adaptive beamforming (SA-RAB)algorithm. According to the symmetry of the space domain and frequency domain, the proposed algorithm uses the spectral analysis technology to get the main beam width based on the error between the ideal steering vector and the real steering vector, and at last uses the second-order cone programming (SOCP) technology to form a flat response in the main beam and suppress the interference in the sidelobe. Simulation result shows that the proposed algorithm can output ideal signal to interference plus noise ratio (SINR) and improve the robustness with the interference being well suppressed when the array steering vector is not accurate, confirming the validity and superiority of the proposed algorithm.