Ocean source localization with multi-frequency deep learning in uncertain environments

Abstract

A deep learning approach is proposed to locate broadband acoustic sources in ocean waveguides with uncertain bottom parameters. The residual neural networks, trained on a huge number of sound field replicas generated by an acoustic propagation model, are used to handle the environmental uncertainty in source localization. A two-step training strategy is presented to improve the training of the deep models. First, the range is discretized in a coarse (5 km) grid. Subsequently, the source range within the selected interval and source depth are discretized on a finer (0.1 km and 2 m) grid. The deep learning methods are demonstrated for simulated magnitude-only multi-frequency data in uncertain environments. Experimental results from the China Yellow Sea show that the approach is comparable with SAGA in performance while much faster in computation speed.