SRLF‐MUSIC: Toeplitz‐Anchored Structure‐Regularised Low‐Risk Covariance Fusion for Low‐SNR Small‐Snapshot MIMO Radar DOA Estimation
Li Che, Lin Li, Liubing Jiang · IET Radar Sonar & Navigation · 2026
ABSTRACT Under low signal‐to‐noise ratio (SNR) and limited‐snapshot conditions, multiple‐input multiple‐output (MIMO) radar direction‐of‐arrival (DOA) estimation is often limited by non‐asymptotic covariance‐matrix instability rather than by the multiple signal classification (MUSIC) pseudospectrum itself. Although the sample covariance matrix preserves finite‐sample information, it can suffer from noise‐eigenvalue spreading, spectral‐gap shrinkage, and subspace leakage under low SNR, few snapshots, and mildly spatially correlated Gaussian noise. Toeplitz projection reduces covariance degrees of freedom by exploiting the virtual uniform linear array (ULA) prior, but fixed structural processing may suppress useful sample‐dependent differences. To address this issue, this paper proposes SRLF‐MUSIC, a Toeplitz‐anchored structure‐regularised low‐risk covariance‐fusion method that preserves the standard MUSIC pseudospectrum and adaptively fuses sample information, structural prior information, and risk‐fallback terms using observable covariance‐risk quantities. Monte Carlo simulations show that SRLF‐MUSIC provides more stable error control than Toeplitz‐MUSIC in the low‐SNR, small‐snapshot threshold region, reducing root mean square error by 33.5% at SNR dB under white Gaussian noise and by 17.7% at SNR dB under spatially correlated Gaussian noise. A RADIal‐derived background case further supports its stability under the measurement‐derived background setting.