Knowledge Aided Covariance Matrix Estimation via Gaussian Kernel Function for Airborne SR-STAP

Mingxin Liu,Xuegang Wang,Xuelian Yu,Lin Zou,Bin Tang,Yun Zhou

doi:10.1109/access.2020.2963838

Abstract

In practical airborne radar, the interference signals in training snapshots usually lead to inaccurate estimation of the clutter covariance matrix (CCM) in space-time adaptive processing (STAP), which seriously degrade radar performance and even occur target self-nulling phenomenon. To solve this problem, a knowledge-aided sparse recovery (SR) STAP algorithm based on Gaussian kernel function is developed. The proposed method distinguishes clutter components and interference signals in training snapshots by the priori knowledge that the clutter components are distributed along the clutter ridge, which dislodges interference signals from training snapshots by Gaussian kernel similar degree. Thus, the CCM is estimated by utilizing these snapshots. Finally, the proposed STAP weight vector is built, which is convenient for the subsequent signal processing. The experimental results were performed to verify the effectiveness and superiority of the proposed method. The test results also show that the proposed algorithm completely removes the interference signals, accurately estimates CCM, and improves the moving target detection performance in small-sample and non-homogeneous clutter environments.

Highlights

Space-time adaptive processing (STAP) can improve the ability of suppressing clutter and detecting targets in airborne radar [1]–[3]
The simulated results show that the proposed method achieves a significant performance improvement for STAP in a small number of training snapshots
In the present study, a knowledge-aided sparse recovery (SR)-STAP approach based on Gaussian kernel similar degree was proposed

Summary

INTRODUCTION

Space-time adaptive processing (STAP) can improve the ability of suppressing clutter and detecting targets in airborne radar [1]–[3]. If the training snapshots contain the interference signals, the CCM estimated by these algorithms is inaccurate, even leading to the target selfnulling effect. In the situation of dense distribution of interference signals or the limited number of the training snapshots, the robustness of the above algorithms is poor These algorithms can degrade the performance of STAP owing to discard many contaminated snapshots in the heterogeneous environment. A knowledge-aided SR-STAP algorithm based on Gaussian kernel function is proposed for clutter suppression and moving target detection in small-snapshots and non-homogeneous clutter environments. The proposed algorithm uses the prior knowledge of clutter spectrum and Gaussian kernel similar degree to detect and eliminate the interference signals in the training snapshots, and accurately estimates the CCM. IM is the M × M identity matrix, and diag(a) denotes a diagonal matrix whose diagonal elements are equal to the VOLUME 8, 2020 column vector a. max(a) denotes the maximum element of vector a. < · > is the inner product

SIGNAL MODEL

GAUSSIAN KERNEL SIMILAR DEGREE

SR TECHNIQUE

THE PROPOSED ALGORITHM

RESULTS AND DISCUSSION

CONCLUSION