DOA Estimation of Coherent Signals From a Single Moving Sensor: Time-Domain Multisampling and Synthetic Aperture Convolution Kernel

Abstract

Direction of arrival (DOA) estimation is a pivotal aspect of array signal processing, particularly in radar and communication systems. In environments characterized by multipath propagation, signal coherence often results in rank deficiency within the data covariance matrix, thereby diminishing the effectiveness of classical subspace methods. Conventional decoherence techniques, such as spatial smoothing, can restore matrix rank but at the cost of reduced array aperture, which in turn constrains source estimation and localization capabilities. Recent advancements in moving array techniques have attempted to address these limitations by generating virtual arrays to enhance the degrees of freedom. However, these methods still experience challenges related to aperture loss or substantial computational demands in coherent scenarios. To overcome these obstacles, this article introduces an approach termed the synthetic aperture (SA) convolution kernel method, which leverages time-domain -sampling technology. The proposed method facilitates high-resolution DOA estimation of coherent signals without incurring aperture loss, utilizing only a single moving sensor. Furthermore, we establish the conditional and unconditional Cramér–Rao bounds (CRBs) for the single-moving-sensor SA model, which provide a theoretical performance benchmark and reveal that the estimation error bound decreases with the square of the sampling factor $\alpha _{i}$. Simulations show that the proposed method outperforms traditional spatial smoothing techniques, achieving better angular resolution and estimation accuracy under low signal-to-noise ratios, closely approaching the derived CRBs.