Sparsity-based Shipping Noise Analysis and Cancellation in Underwater Acoustic Communication
Ruoyu Zhang,Lutz Lampe,Honglin Zhao
DOI: https://doi.org/10.1121/1.5067682
2018-01-01
Abstract:We are interested in achieving reliable underwater acoustic communication (UWAC) in the presence of noise generated by vessel traffic, i.e., shipping noise. Our general approach is to determine an underlying structure in shipping noise and to exploit this structure to remove it from desired UWAC signals. Since a well-established statistical characterization of shipping noise is not available, in this paper, we discuss our evaluation of shipping noise data sets obtained from labeled recordings from an underwater acoustic hydrophone array deployed in the Strait of Georgia by Ocean Networks Canada. Motivated by the general observation that shipping noise exhibits a structure akin to impulse noise, we study to what extent shipping noise can be described as a sparse signal. Specifically, we apply the expectation-maximization method to train a Gaussian-mixture model with two components, in which the large variance component represents the impulse-noise part. In a second step, we use the model to perform noise cancellation in a multicarrier UWAC scenario. For this purpose, we apply a sparse-signal recovery formulation and a message passing algorithm for detection, i.e., a form of compressed sensing. Numerical results show the moderate level of mitigation of shipping noise interference achieved with the proposed method.We are interested in achieving reliable underwater acoustic communication (UWAC) in the presence of noise generated by vessel traffic, i.e., shipping noise. Our general approach is to determine an underlying structure in shipping noise and to exploit this structure to remove it from desired UWAC signals. Since a well-established statistical characterization of shipping noise is not available, in this paper, we discuss our evaluation of shipping noise data sets obtained from labeled recordings from an underwater acoustic hydrophone array deployed in the Strait of Georgia by Ocean Networks Canada. Motivated by the general observation that shipping noise exhibits a structure akin to impulse noise, we study to what extent shipping noise can be described as a sparse signal. Specifically, we apply the expectation-maximization method to train a Gaussian-mixture model with two components, in which the large variance component represents the impulse-noise part. In a second step, we use the model to perform noise c...