Nonlinear internal waves have been shown to distort acoustic reflection measurements such as those obtained from a sub-bottom profiler. For moored transceivers, the reflection measurements can be modeled as sparse signals with the same support, where the non-zero elements correspond to reflections off the seabed layers. Nonlinear internal waves thus have the effect of altering the acoustic propagation channel, acting as a filter applied to this sparse signal. We model the problem of recovering the bottom reflections as one of sparse blind deconvolution with multiple measurement vectors. We propose an alternating method that jointly recovers the sparse bottom reflections and the unknown time-varying filters. Empirical results show that the proposed method converges to a local minimum when the regularization parameters are selected appropriately. We evaluate our method on data from two moorings in the South China Sea, one experiencing significant NIWs. Applying our method shows that we are able to mitigate the focusing and defocusing effects of NIWs and robustly estimate the bottom and sub-bottom reflection times. In addition, examination of the learned filters shows that NIWs have the effect of blurring and time-shifting the true signals. Work supported by the Office of Naval Research.
Lipor et al. (Wed,) studied this question.