Abstract: Traditional adaptive beamformer suffers from performance degradation in the presence of mismatch between actual and assumed array steering vectors of the desired signal.A novel approach to adaptive beamformer is proposed,which emphasizes the robustness to the system errors.Firstly,the desired signal component is attenuated from input snapshot data by spatial matrix filter;then the covariance matrix is reconstructed by using the filtered snapshot data and transformed back to element level data prior to performing adaptive beamforming;finally,the weights of the adaptive beamformer are computed using transformed covariance matrix.The method overcomes the desired signal cancellation problem of adaptive beamformer caused by system errors because the desired signal component is attenuated by spatial matrix filter.Therefore,the robustness of adaptive beamformer against the system errors is greatly improved.Computer simulation results prove its validity and superiority to the existing methods.