2025_programme: Source Localization in Shallow Water Using a Neural Network Based on Range-Dependent Sound Speed Profile Model
- Day: June 17, Tuesday
Location / Time: B. ERATO at 11:00-11:20
- Last minutes changes: -
- Session: 19. Uncertainty quantification and machine learning in signal processing
Organiser(s): Angeliki Xenaki, Zoi-Heleni Michalopoulou
Chairperson(s): Angeliki Xenaki, Zoi-Heleni Michalopoulou
- Lecture: Source Localization in Shallow Water Using a Neural Network Based on Range-Dependent Sound Speed Profile Model [Invited]
Paper ID: 2144
Author(s): Jing Guo, Juan Zeng, Li Ma
Presenter: Jing Guo
Abstract: In range-dependent shallow-water environments, the complex spatiotemporal variations of the environmental sound speed profile (SSP) lead to mismatches between the SSP model and the actual environment, posing challenges in passive source localization. Leveraging the nonlinear feature extraction and model fitting capabilities of deep learning techniques, this paper investigates the impact of range-dependent SSP on source localization. The performance of a deep neural network (DNN) trained on a simple equivalent SSP model is evaluated, where the environmental SSP is approximately parameterized as a function of distance. The test results indicate that the neural network, after training on the dataset generated from the equivalent SSP model, demonstrates the ability to localize sources in range-dependent shallow-water environments, albeit with relatively large errors. By applying transfer learning with experimental data from the Yellow Sea, the range-dependent network better fits the environmental model, leading to reduced errors and improved localization accuracy.
Download the full paper
This paper is a candidate for the This paper is a candidate for the "Prof. Leif Bjørnø Best Student Paper Award (for students under 35)"
- Corresponding author: Dr Jing Guo
Affiliation: Key Laboratory of Underwater Acoustic Environment, Institute of Acoustics, Chinese Academy of Sciences, China
Country: China
