Robust adaptive wideband beamforming for passive acoustic array under impulsive noise

Robust wideband beamforming is critical for passive acoustic sensing systems operating in challenging underwater environments characterized by impulsive noise, interference, and array uncertainties. Conventional robust wideband beamforming methods are generally developed under the Gaussian noise assumption and rely on second-order sample covariance matrix estimation. In impulsive underwater acoustic environments, high-amplitude outliers severely corrupt the sample covariance matrix, degrading steering-vector estimation, interference-plus-noise covariance matrix (INCM) reconstruction, and ultimately the overall beamforming performance. To address this limitation, this paper proposes a robust wideband beamforming framework in which a fractional lower-order moment-based time-frequency (TF) spatial covariance matrix provides a statistically robust covariance estimate for the subsequent steering-vector refinement and improved Capon-based INCM reconstruction. The proposed method is suitable for wideband, non-stationary signals and maintains reliable covariance estimation under limited-snapshot conditions through TF domain processing and covariance averaging. Numerical simulations demonstrate improved output SINR under impulsive noise, direction-of-arrival mismatch, sensor perturbations, and limited-snapshot scenarios. SWellEx-96 experimental results further demonstrate improved suppression of interference and sidelobe energy.