Graph-based sequential beamforming.

Area of Science:

• Signal Processing
• Statistical Inference
• Array Signal Processing

Background:

• Direction finding is crucial in various applications like radar and sonar.
• Sequential estimation methods are needed for dynamic environments with time-varying signals.
• Existing methods may struggle with accuracy in complex, time-varying scenarios.

Purpose of the Study:

• To develop a Bayesian estimation method for sequential direction finding.
• To accurately estimate the number and locations of directions of arrival (DOAs) over time.
• To improve upon nonsequential and state-of-the-art DOA estimation techniques.

Main Methods:

• Utilizes a factor graph to model sequential beamforming.
• Employs belief propagation for predicting DOAs at each time step.
• Applies Variational Bayesian inference for updating DOA estimates.
• Incorporates a Bernoulli-Gaussian amplitude model for sparsity and a gridless approach.

Main Results:

• The method effectively estimates the number and locations of DOAs.
• Achieves reduced DOA estimation errors in scenarios with multiple time steps and time-varying DOAs.
• Demonstrates performance improvements over nonsequential and state-of-the-art methods in simulations.
• Provides marginal posterior probability density functions for uncertainty quantification.

Conclusions:

• The proposed Bayesian sequential direction finding method offers enhanced accuracy and robustness.
• It is suitable for applications with dynamic and multiple sources.
• Validated through simulations and ocean acoustic experimental data.