Separation of moving ship striation patterns using physics-based filtering
Résumé
When a ship is moving toward an acoustic receiver in an oceanic waveguide, the time-frequency representation of the recorded signal exhibits a striation pattern that can be useful in numerous applications such as ship localization or geoacoustic inversion. If there are many ships, the striation patterns add up and they must be separated if one wants to study them separately. In this paper, a physics-based filtering scheme for passive underwater acoustics has been developed. The algorithm allows separating the time-frequency striations of two different moving ships. The proposed method considers filtering the 2D Fourier transform of the received spectrogram. The filter design is based on the waveguide invariant principle and on some a priori knowledge on the oceanic waveguide. The noise nature on the spectrogram is taken into account by introducing a nonlinearity to the filtering scheme. The algorithm thus corresponds to a nonlinear homomorphic filter. The method is validated on both simulated data and experimental marine data. This filtering scheme offers good prospects for all applications using ship noise and a single receiver.