Multi-layer residual beamforming for underwater acoustic source bearing estimation

Area of Science:

• Acoustics
• Signal Processing
• Machine Learning

Background:

• Deconvolution beamforming offers robustness in underwater source localization against array mismatch and limited snapshots.
• However, existing methods face limitations in resolution and accuracy for precise source identification.

Purpose of the Study:

• To develop an advanced beamforming framework for enhanced underwater source localization.
• To improve the resolution and accuracy of direction-of-arrival estimation in challenging acoustic environments.

Main Methods:

• A multi-layer residual beamforming framework inspired by deep learning architectures was proposed.
• The framework utilizes 2K layers of alternating phase compensation and source cancellation for K source localization.
• Methods were validated through simulations and experimental data from the SWellEx-96 dataset.

Main Results:

• The proposed method demonstrated robustness against array mismatch and limited snapshots.
• It achieved estimation accuracy comparable to the least absolute shrinkage and selection operator (LASSO).
• The framework significantly outperformed traditional deconvolution beamforming in simulations.

Conclusions:

• The multi-layer residual beamforming framework offers a practical and effective solution for underwater source localization.
• The method provides high accuracy and robustness, addressing limitations of existing techniques.
• Experimental validation confirms its feasibility for real-world acoustic applications.