Machine learning based ship localization in shallow water using ship noise recorded by two vertical line arrays

For the application of machine learning to sound source localization, much train data distinguished from test data is needed to build the machine learning model. In Shallow-water Acoustic Variability Experiment (SAVEX-15) conducted in shallow water (water depth ∼100 m) in Northern East China Sea (ECS), ship noise of the R/V Onnuri was recorded by two vertical line arrays. Acoustic data of ∼80% was applied to the training dataset and the others having different trajectories were used for the test data. The recorded data is preprocessed by a sample covariance matrix and it is used as the input data of the machine learning model: Feedforward neural network (FNN) and support vector machine (SVM). The results by FNN and SVM will be discussed with conventional localization method using ray-based blind deconvolution (RBD) and array invariant (AI). [Work supported by the Agency for Defense Development, Korea (UD210004DD).]