Wideband DOA Estimation Based on Stochastic Maximum Likelihood Estimation With Flat Spectra Assumption

This paper introduces a novel method for wideband direction-of-arrival estimation within the stochastic maximum likelihood framework. The flat power spectra assumption is leveraged, which implies consistent signal statistics across subbands. This assumption allows for convex constraints to be imposed on the covariance matrices, leading to a simplified optimization. The proposed approach demonstrates robustness to deviations from the flat spectra assumption, achieving asymptotically unbiased estimates at high signal-to-noise ratios. Additionally, the method guarantees convergence to a local minimum, thanks to its formulation using the majorization-minimization principle and the avoidance of non-convex constraints. Simulations validate the effectiveness and stable performance of the proposed method, which demonstrates satisfactory performance even when the number of data snapshots is limited.