Spatial Power Spectrum Estimation Under Strong Interferences Using Beam-Space Fast Nonnegative Sparse Bayesian Learning

In acoustic array signal processing, spatial power spectrum estimation and the associated direction-of-arrival (DOA) estimation are often inflicted by strong interferences, which lead to significant performance degradation and even mask the weak targets. Although tremendous efforts have been put into related research, simultaneously realizing robust interference suppression and DOA estimation against various model mismatches is still challenging. To address this challenge, this article proposes a systematic scheme that cohesively integrates a robust beamforming step and a beam-space sparse learning step, to effectively recover the spatial power spectrum in the presence of strong interference. Considering the sparsity and nonnegativity of the spatial power spectrum, we propose a nonnegative fast sparse Bayesian learning algorithm to reconstruct the spatial power spectrum of target sources from the beam-space data. In addition to the outstanding interference suppression capabilities, our method exhibits better denoising performance (i.e., lower noise level) and DOA estimation accuracy, even under challenging scenarios, such as snapshot deficiency and low signal-to-noise ratios. Simulated and real-life experimental data results verify the robustness and superior performance of the proposed scheme over other competitors.