In this work, we propose a new diagonal loading approach to improve the robustness of the standard Capon beamformer. For this purpose, the array is analyzed using spatial autoregressive modeling. The proposed regularization method is orderrecursive, which allows the definition of a Levinson-like algorithm. This eliminates the need to invert the regularized array covariance matrix, essential for all types of Capon beamformers. The proposed diagonal-loading approach involves two regularization parameters, thereby increasing the robustness compared to one-parameter diagonal-loading approaches. © 2011 IEEE.