Convolutional Neural Network-Based Underwater Acoustic Source Localization

Passive localization of underwater acoustic sources plays a critical role in underwater target detection and identification. To achieve accurate and robust passive localization, this study leverages underwater acoustic data and applies a convolutional neural network (CNN) to the task of underwater source localization. Specifically, a multi-task convolutional neural network (MTL-CNN) framework is proposed to jointly estimate the range and depth of the acoustic source. A novel network architecture, referred to as MTL-CNN-r, is developed to support this joint localization task. The proposed method is evaluated using both a simulation dataset generated based on measured ocean environment parameters and the publicly available SWellEx-96 Event S5 dataset. Experimental results demonstrate that the MTL-CNN-r model is capable of simultaneously estimating the range and depth of the acoustic source under unknown ocean environmental parameters, with low localization errors.