Dual-branch time–frequency domain anti-interference method for ship radiated noise signal

Abstract

Sensing the target ship using the radiated noise of the ship is the main means of sensing the target ship. In the adversarial and non-cooperative scenarios, explosion interference is a typical interference source used to mask the target ship and confuse the underwater acoustic sensing system to effectively perceive the target. In this paper, we try to use the idea of signal enhancement to deal with the anti-explosion interference. we try to enhance the ship radiated noise signal with extremely low signal-to-noise ratio, so as to achieve anti-explosion signal interference. We propose a deep learning approach with encoder–decoder as the basic framework. We design a decoder with two branches, one for estimating the he mask of the noise signal amplitude spectrogram and the other for estimating the complex spectrogram. We add the attention mechanism to the model to select features relevant to each branch task. Finally, we use the real experimental data to test, the experimental results show that our method can enhance the ship radiation noise signal to 2 dB under the condition of −20 dB to −25 dB very low signal-to-noise ratio. The experimental results fully prove that our method can achieve anti-explosion interference of ship radiated noise.