Source depth discrimination based on deep learning using frequency-wavenumber domain features

Source depth discrimination using a horizontal array is a challenging task. This paper addresses the discrimination of source depths by exploiting modal excitation differences of surface and underwater targets in a shallow water environment. The modal domain beamforming technique is utilized to estimate the wavenumber spectrum. Then frequency-wavenumber domain features are extracted as inputs of a convolutional neural network (CNN). The CNN is acted as a binary classifier for target depth estimation. The simulation experimental results have shown that the proposed method outperforms the conventional matched field processing.