Mode-informed complex-valued neural processes for matched field processing.

Area of Science:

• Ocean acoustics
• Signal processing
• Machine learning

Background:

• Accurate acoustic field reconstruction is crucial for underwater passive source localization.
• Traditional methods face challenges with sparse data and computational efficiency.

Purpose of the Study:

• To develop an advanced method for acoustic field reconstruction using complex-valued neural processes and modal depth functions (MDFs).
• To improve the accuracy and robustness of passive source localization in ocean environments.

Main Methods:

• Utilizing complex-valued neural networks to model acoustic field distributions with Gaussian processes.
• Employing a meta-learning strategy for parameter optimization to prevent overfitting and accelerate reconstruction.
• Integrating the reconstructed acoustic field with matched field processing (MFP) for localization.

Main Results:

• The proposed method significantly improves localization performance on the SWellEx-96 waveguide.
• The MDF-based kernel demonstrates superior performance over the Gaussian kernel in characterizing ocean waveguides.
• The method effectively denoises, interpolates sparse data, and generates dense field data for virtual arrays.

Conclusions:

• The complex-valued neural process method with MDFs offers enhanced acoustic field prediction and robust passive source localization.
• This approach addresses limitations of traditional MFP and GP-based MFP, particularly in complex ocean environments.