Abstract: The performance of Direction Of Arrival(DOA)estimation based on Compressed Sensing(CS)decreases in the complex ocean marine environment.In order to tackle this problem,a method of DOA estimation for underwater acoustic target based on CS after blind reconstruction of array signal in frequency domain is proposed.Firstly,the received array data are transformed to frequency domain by Fourier transform and frequency domain wide band signal are divided into part overlapping multiple sub-band array signal.Secondly,each sub-band array signal are separated using plural Blind Source Separation(BSS)method,the sub-band separated matrix and target signal can be estimated.Thirdly,the array signal in frequency domain are reconstructed according to the separated matrix and separated signals which were not noises.Fourthly,the sub-band spatial spectrum corresponding to the reconstructed array signal are got by CS beamforming method.Finally,the total spatial spectrum can be got by summing the all sub-band spatial spectrum.And the target direction can be estimated by searching the peak value of the total spatial spectrum.The verification results of simulator data and sea measured data show that,under the same conditions,the target detection ability and direction precision of the proposed method is superior to the classical Minimum Variance Distortionless Response(RIVDR)method,frequency domain CS method,BSS combined with MVDR method.The spatial spectrum energy of faint target signal is improved obviously,and the ability of the sonar to detect faint target is enhanced.