Source Localization Using Gradient Boosting Decision Tree with a Single Hydrophone in Deep Ocean

Abstract

In deep ocean environment, the horizontal range of the direct arrival zone is far when the receiver is near-bottom and the source is near-surface. The transmission loss in the direct arrival zone is less than that in shadow zone, which is beneficial for underwater sound source detection. This paper proposes the use of gradient boosting decision tree (GBDT) for source localization in deep ocean with a single hydrophone. The input features are multi-path time delays calculated by an acoustic propagation model or extracted from the autocorrelation function. The observed data from the deep ocean area of the South China Sea (SCS) are used to validate the proposed method. The results verify the performance of the method to localize a near-surface broadband source, using a near-bottom single hydrophone in deep ocean.