Improved Variational Bayes for Space-Time Adaptive Processing

Area of Science:

• Signal Processing
• Radar Systems
• Statistical Inference

Background:

• Moving target detection in radar faces challenges with small sample sizes and non-uniform environments.
• Sparse signal reconstruction offers a feasible approach for estimating clutter spectra in the angle-Doppler domain.
• Existing Sparse Bayesian Learning (SBL) methods show limitations in enhancing sparsity and robustness.

Purpose of the Study:

• To improve the sparsity and robustness of Space-Time Adaptive Processing (STAP) algorithms for moving target detection.
• To address the computational complexity associated with variational inference techniques in sparse recovery.
• To enhance the accuracy and efficiency of clutter spectrum estimation in challenging environments.

Main Methods:

• Introduction of a hierarchical Bayesian prior framework with iterative parameter updates via variational inference.
• Development of an enhanced Variational Bayesian Inference (VBI) method utilizing prior rank information of the temporal clutter covariance matrix.
• Application of the Multiple Measurement Vector (MMV) model for joint sparsity and a first-order Taylor expansion to mitigate dictionary grid mismatch.

Main Results:

• Significantly improved sparsity and robustness in clutter spectrum estimation.
• Substantial reduction in computational complexity for parameter updates.
• Enhanced moving target detection performance in complex, non-uniform environments with limited data.

Conclusions:

• The proposed enhanced VBI method effectively improves sparsity and reduces computational load in STAP algorithms.
• This research offers novel methodologies for sparse signal reconstruction in radar signal processing.
• The findings contribute to more robust and accurate moving target detection capabilities.