Target depth estimation by deep neural network based on acoustic interference structure in deep water

Automatic and robust target depth estimation is an important issue of active detection system working in the reliable acoustic path (RAP) environment. In this paper, depth-sensitive interference structure used as input, a deep neural network (DNN) method proposed to realize automatic estimation. Furthermore, improve robustness network, both sound speed profile (SSP) uncertainty trajectory are considered training dataset. The former modelled random weighted sum empirical orthogonal function (EOF), latter by partial-data-lacking input. was evaluated through simulated data compared with conventional matching measured replicas. presented exhibits higher accuracy estimate uncertain environments when signal-to-noise ratio (SNR) than ?6 dB, but its performance deteriorates rapidly SNR decreases further. also verified semi-simulation experimental water.