Off-grid Compressive Matched Field Processing

‍ ‍Passive target localization in underwater acoustics plays a crucial role in tasks such as marine resource exploration， underwater detection， and maritime security monitoring. It stands as one of the most pivotal subjects within the underwater acoustics domain. Proposed methods such as conventional matched field processing （MFP） and compressive MFP （CMFP） can effectively localize sources at the pre-divided range-depth grids. However， deviations of actual target sources from these grid points lead to performance degradation in such methods. In addition， the off-grid source may lead to multiple sparse solutions corresponding to a single source in CMFP， resulting in multi-peak ambiguity. Addressing these challenges， considering that the channel impulse response （CIR） at adjacent underwater locations present significant similarities， this paper capitalizes on the property that the CIR at any off-grid location can be approximated by the replica vectors at neighboring grid points. Based on this property， we established a group sparse signal model oriented towards underwater target localization. Subsequently， a corresponding off-grid compressive matched field processing （OG-CMFP） method is proposed. This method leverages the recovered sparse signal to compute interpolation coefficients at grid points， enabling high-precision localization of off-grid sources. Experimental results demonstrate that， OG-CMFP outperforms traditional MFP and CMFP techniques in a two-source localization task. With a range interval of 0.05 km and a depth interval of 10 m， under conditions of SNR greater than -5 dB， OG-CMFP achieves an average rMAE reduction of over 250 m compared to CMFP.