PaperlayerView exact revision on arXiv

Source-linked AI summary

Direct Data Domain STAP using Sparse Representation of Clutter Spectrum

Ke Sun, Huadong Meng, Yongliang Wang, Xiqin Wang

arXiv:1008.4184v1cs.IT

TL;DR

Non-stationary clutter and non-side-looking configurations undermine training-data-based STAP, while direct data-domain methods trade training-data independence for reduced DOF. The paper proposes D3SR, which estimates a high-resolution spectrum from the test cell using sparse representation and achieves better output SCR and MDV than current D3 methods in both side-looking and non-side-looking cases.

  • Problem

    Fast-changing clutter or non-side-looking configurations destroy training-data stationarity, degrading statistical STAP, while existing direct data-domain methods reduce system DOF.

  • Method

    D3SR exploits prior sparsity in the test-cell spectral distribution to estimate a high-resolution space-time spectrum, then obtains the CCM and adaptive filter without training data.

  • Results

    D3SR provides better output SCR and MDV than current direct data-domain methods in both side-looking and non-side-looking cases.

  • Takeaways & Limitations

    Maintaining full system DOF enables D3SR to address nonstationarity while improving the reported SCR and MDV performance.

  • Takeaways & Limitations

    The method's sparsity assumption may weaken when the designed spectrum differs substantially from the actual spectrum, a condition noted as common in short-range cases.

Abstract

from arXiv · show

Space-time adaptive processing (STAP) is an effective tool for detecting a moving target in the airborne radar system. Due to the fast-changing clutter scenario and/or non side-looking configuration, the stationarity of the training data is destroyed such that the statistical-based methods suffer performance degradation. Direct data domain (D3) methods avoid non-stationary training data and can effectively suppress the clutter within the test cell. However, this benefit comes at the cost of a reduced system degree of freedom (DOF), which results in performance loss. In this paper, by exploiting the intrinsic sparsity of the spectral distribution, a new direct data domain approach using sparse representation (D3SR) is proposed, which seeks to estimate the high-resolution space-time spectrum with only the test cell. The simulation of both side-looking and non side-looking cases has illustrated the effectiveness of the D3SR spectrum estimation using focal underdetermined system solution (FOCUSS) and norm minimization. Then the clutter covariance matrix (CCM) and the corresponding adaptive filter can be effectively obtained. Since D3SR maintains the full system DOF, it can achieve better performance of output signal-clutter-ratio (SCR) and minimum detectable velocity (MDV) than current D3 methods, e.g., direct data domain least squares (D3LS). Thus D3SR is more effective against the range-dependent clutter and interference in the non-stationary clutter scenario.

2 Science and Research Department, Wuhan Radar Academy, Wuhan 430019, China)

The paper concerns STAP using sparse representation for non-stationary radar clutter without training-data or DOF-loss concerns.

  • STAP and sparse representation are central topics.
  • The approach emphasizes no training data and no DOF loss.
  • FOCUSS is associated with non-stationary clutter processing.

1. Introduction

STAP must estimate clutter accurately to detect moving targets, but nonstationary clutter can undermine training-based methods, while D3 methods trade training-data independence for reduced system DOF. The proposed D3SR method uses sparse spectral representation from only the test cell to retain higher DOF and improve clutter suppression and detection performance.

  • Motivation: STAP detects moving targets in Doppler/angle-spread clutter, where joint space-time processing and adaptive filtering are needed to distinguish targets from surrounding clutter.Accurate clutter-spectrum knowledge supports construction of an adaptive filter that improves output SCR.
  • Limitations of existing methods: Training-based methods require IID data for effective CCM estimation, but nonhomogeneous or non-side-looking configurations destroy range stationarity and can cause improper clutter nulling.These methods can improve output SCR when sufficient IID training data are available.
  • Direct data domain processing: D3 methods avoid nonstationary training data and suppress clutter and interference in the test cell, but their reduced system DOF causes decreased performance.The tradeoff is between adaptive-filter DOF and the number of subarrays.
  • Proposed method: D3SR estimates a high-resolution space-time clutter spectrum from only the test cell by formulating spectral estimation as a sparse-constrained underdetermined inverse problem.The approach exploits prior sparsity in the test-cell spectral distribution and uses sparse-representation algorithms such as FOCUSS and norm minimization.
  • Proposed method: D3SR extracts clutter distribution from the estimated spectrum, constructs the CCM and adaptive filter, and provides higher system DOF than D3LS in side-looking and non-side-looking cases.The paper reports better output SCR and MDV performance and effectiveness against nonstationary clutter.

2. Signal Model

The signal model describes nonstationary airborne STAP data containing a moving target, clutter, discrete interference, and noise. It explains how array geometry creates range-dependent clutter ridges and motivates direct-data-domain filtering, whose training-data robustness trades off against reduced system degrees of freedom.

  • Signal and clutter characteristics: Airborne STAP data are nonstationary because array configuration, flight geometry, and changing clutter conditions alter the clutter distribution across range cells.In non-side-looking configurations, Doppler depends on look direction and range, producing range-dependent clutter ridges.
  • Signal and clutter characteristics: In side-looking geometry, clutter ridges from different range cells coincide as a straight angle-Doppler line, whereas non-side-looking geometry produces concentric elliptical trajectories.The resulting range dependence becomes especially serious at short range and can make training-based covariance estimates produce improper filters.
  • Signal and clutter characteristics: The received test-cell data comprise a moving target, terrain clutter, discrete interferers, and thermal noise.Clutter is modeled as independent scatterers, while discrete interferers are modeled separately and need not lie on the clutter ridge.
  • STAP objective: STAP seeks to suppress clutter and interference while preserving signal-of-interest gain to improve output signal-clutter ratio.Accurate knowledge of the test-cell clutter ridge is required for high-resolution Doppler separation of moving targets from clutter.
  • Direct-data-domain processing: Direct-data-domain least squares uses only the test cell, avoiding stationary training-data requirements and suppressing clutter and discrete interferers.It cancels the possible signal of interest through a subarray construction and solves for an adaptive filter using least squares.
  • Direct-data-domain processing: The direct-data-domain benefit is offset by reduced system degrees of freedom, which can cause performance loss and motivates a new method for nonstationary clutter.The stated reduction is N_aP N_a < NM, and the proposed approach targets this remaining limitation.

3. SPACE-TIME SPECTRUM ESTIMATION

D3SR estimates a high-resolution space-time spectrum from only the test cell by exploiting sparsity in the angle-Doppler distribution. This avoids nonstationary training data while preserving system degrees of freedom for subsequent clutter suppression.

  • Statistical training-data stationarity is destroyed in fast-changing scenarios or non-side-looking configurations, limiting conventional clutter-spectrum estimation.
  • D3SR estimates a high-resolution spectrum using only the test cell, avoiding the need for stationary training data.
  • D3SR uses the estimated spectrum to extract clutter and interference, obtain the clutter covariance matrix, and construct an adaptive filter.
  • 3. SPACE-TIME SPECTRUM ESTIMATION: D3SR discretizes angle and Doppler axes and represents the test-cell data with an overcomplete basis of space-time steering vectors.
  • 3.2 Spectrum estimation by sparse representation: Sparsity constrains the ill-posed representation problem, while 1L norm minimization and FOCUSS provide practical sparse-solution algorithms.
  • 3.1 The sparsity of spectral distribution: The angle-Doppler test-cell distribution is sparse because clutter, interferers, and possible targets occupy relatively few cells.
  • 3.2 Spectrum estimation by sparse representation: Adaptive FOCUSS iteratively reinforces prominent spectral entries and suppresses others, potentially improving sparse representation and noise robustness.

4. EXPERIMENTAL RESULTS

Simulations evaluate sparse spectrum estimation and adaptive filtering in side-looking and non-side-looking airborne-radar scenarios. D3SR suppresses clutter and interference while retaining full system degrees of freedom, yielding better output behavior than D3LS in supported cases.

  • The experiments cover side-looking and non-side-looking airborne-radar arrays, including stationary-disrupting interference or range-dependent clutter.
  • 4.1 Spectrum estimation using sparse representation: Both 1L norm minimization and adaptive FOCUSS obtain high-resolution spectra and reduce clutter or interference spread in the side-looking case.
  • 4.1 Spectrum estimation using sparse representation: In the non-side-looking case, both sparse estimators obtain high-resolution spectra and decrease clutter spread, while the target remains detectable in the estimation result.
  • 4.1 Spectrum estimation using sparse representation: Adaptive FOCUSS provides better spectrum estimation and lower computational load than 1L norm minimization, so D3SR adopts its estimate for filtering.
  • LSMI leaves interference because it is absent from training data, whereas D3LS and D3SR suppress clutter and interference and improve output SCR.
  • Both D3LS and D3SR degrade near the clutter notch, but D3SR provides better output SCR when the target is away from that notch.
  • D3SR has less clutter residual, a narrower clutter notch, and better pass-band SCR improvement than D3LS because it retains full system degrees of freedom.

5. CONCLUSION

The paper proposes D3SR, a direct-data-domain method that exploits spectral sparsity to estimate high-resolution spectra in non-stationary clutter scenarios. It estimates the test-cell CCM accurately, preserves full system DOF, and improves output SCR and MDV, while future extensions require reconsidering the dictionary and sparsity assumptions.

  • D3SR addresses non-stationary clutter in both side-looking and non-side-looking configurations.
  • D3SR uses sparse representation, including 1L norm minimization and adaptive FOCUSS, to obtain high-resolution spectral estimates.
  • Accurate test-cell CCM estimation enables an effective adaptive filter while retaining the full system DOF.
  • D3SR achieves better output SCR and MDV performance than the compared direct-data-domain methods.
  • Further research: The fixed overcomplete dictionary may mismatch practical data, reducing sparsity when clutter motion or channel mismatch is present.
  • Further research: Bistatic or conformal-array extensions require reconsidering the dictionary and sparsity, with adaptive sparse-representation mechanisms needed for reliable estimation.
Loading 1008.4184v1…