Underwater target detection using a pair of omnidirectional transducers in extremely shallow water

Abstract

In extremely shallow water, such as bays and inlets, active sonar may not work sufficiently well to detect underwater threats owing to multiple reflections and strong reverberations at the sea surface and seafloor. In this study, we developed "gate sonar" on the basis of a unique method of detecting changes in the pulse signal propagated between a pair of omnidirectional transducers in a surveyed area when a target passes between the transducers. To detect changes in the received signal, we employed a time-reversed pulse as the source signal for the gate sonar; this pulse can be created even from a pair of transducers by using a virtual array formed owing to reflections at the surface and bottom boundaries. As a result of experiments in a 25 m-long swimming pool, we confirmed that the gate sonar can recognize the changes in the received signal and functions as designed. Furthermore, by applying convolutional neural network (CNN) to the received signal waveform and its spectrogram as input images, we demonstrated that the proposed time reversal mirror (TRM)-based gate sonar can improve the detection of targets, with the spectrogram performing slightly better.