Waveguide invariant distribution estimation using autoregressive - Discrete Fourier transform and its application in source depth classification

In shallow water with a thermocline, the waveguide invariant (WI) is typically characterized as a distribution largely dependent on the source depth. To discriminate the source depth under conditions of limited horizontal array aperture and low-frequency broadband sources, an autoregressive-discrete Fourier transform (AR-DFT) method is proposed for extracting the WI distribution. This method combines discrete Fourier transform with an autoregressive (AR) model to estimate the two-dimensional power spectrum, and the WI distribution is extracted based on its relationship with the polar angle of the two-dimensional spectrum ridge. Taking advantage of the characteristic that AR spectral resolution is less affected by aperture size, the AR-DFT method can reduce aperture requirements compared to the traditional two-dimensional discrete Fourier transform (2D-DFT). Subsequently, the dependence of the WI distribution on the source depth in a thermocline waveguide is analyzed. The differences in WI distribution between submerged and surface sources allow the peak position, extracted by the AR-DFT method, to serve as a decision metric for binary source depth classification under small aperture conditions. Simulation and sea trial results confirm the feasibility and effectiveness of the AR-DFT classifier. In comparison with the traditional 2D-DFT classifier, the AR-DFT method demonstrates greater adaptability to smaller aperture arrays. •A WI distribution estimation method to improve the aperture requirement of existing 2D-DFT and its application in source depth discrimination.•The discrete Fourier transform and autoregressive spectrum joint estimation.•Good performance in the simulation process and practical applications.