Reverberation suppression for detecting underwater moving target based on robust autoencoder

Abstract

Reverberation constitutes the main background interference in underwater target detection, increasing the difficulty of extracting target echoes, particularly for moving targets. A robust autoencoder-based method was proposed to suppress reverberation and facilitate moving target detection. The overall received data were decomposed into reverberation, ambient noise, and target echoes. Robust autoencoder utilized a nonlinear representation, which can adaptively estimate and suppress the reverberation to effectively extract the target echoes. Connected component analysis was used to highlight the trajectory of the target by splicing the echoes over continuous pings. Simulations were conducted to test the reliability of the robust autoencoder in highlighting target echoes, which was extended to analyze echoes of a moving target from field experiments using an echosounder with an operating frequency of 38 kHz. Both simulation and field experimental results showed that the proposed method can effectively suppress the reverberation and extract clean target echoes.