Block sparse Bayesian learning with environmental perturbation for robust matched field processing

Compressive sensing (CS) theory offers a transformative approach to overcoming the inherent limitations of matched field processing (MFP) techniques. However, conventional CS-MFP methods usually rely on fixed prior model parameters to construct a dictionary matrix, which often fails to effectively address environmental mismatches caused by dynamic variations and parameter uncertainties in the ocean. To improve the environmental robustness of CS-MFP, a block dictionary matrix with multiple constraints is constructed using the first- and second-order statistics of random environmental parameters. On this basis, a multisnapshot block signal processing model incorporating environmental perturbations is proposed, transforming the matched field localization problem into a block sparse signal recovery problem. Within this framework, an efficient multisnapshot block sparse Bayesian learning (BSBL) algorithm is derived. Furthermore, by establishing different intrablock correlation models, two BSBL processors with enhanced robustness against environmental mismatch are developed. The effectiveness of the proposed method was validated through numerical simulations and real data collected north of Elba in 1993. The results indicate that, compared with the conventional sparse Bayesian learning processor, both proposed BSBL processors do not rely on specific prior environmental parameters and exhibit stronger robustness under environmentally mismatched conditions.