Noise source localization and identification using frequency-difference generalized inverse beamforming

Abstract

Localizing underwater target noise sources to identify the spatial positions of dominant contributors is a prerequisite for effective noise control. However, the noise generated by underwater targets is typically a mixture of broadband and linespectrum signals. Due to array sparseness when processing out-of-band (high-frequency) signals, conventional in-band beamforming (CBF) and generalized inverse beamforming (GIB) often suffer from spatial aliasing, which causes ghost sources. Based on the concept of frequency-difference beamforming (FDB), this paper proposes a frequency-difference generalized inverse beamforming method (FD-GIB). The proposed method first performs frequency-difference processing on the broadband signal to obtain the equivalent array output matrix at the difference frequency Δf. Secondly, a new covariance matrix is constructed based on this matrix and subjected to eigenvalue decomposition, where a truncation threshold is applied to suppress the influence of noise. Then, the acoustic transfer function of the frequency-difference equivalent matrix is constructed, and Tikhonov regularization is introduced, together with an ℓ2-norm constraint for solving. Finally, simulations and experiments demonstrate that FD-GIB suppresses spatial aliasing when processing out-of-band signals, improves source identification accuracy, and achieves marginally superior resolution compared to FDB.