Non-Coherent DOA Estimation With Enhanced Signal Identifiability in Partly Calibrated Arrays

Abstract

This letter addresses the limited signal identifiability in non-coherent direction-of-arrival (DOA) estimation for partly calibrated arrays (PCAs) consisting of identical subarrays. In non-coherent processing, we obtain distinct phase offsets from subarrays of different structures. This allows the identification of more signals than a single subarray can achieve. However, when all subarrays share an identical array geometrical structure, the number of available phase offsets remains constrained to that of a single subarray. To improve the signal identifiability, we propose a non-coherent DOA estimation algorithm with enhanced signal identifiability (called E-MUSIC). In this algorithm, a row-full-rank pseudo-signal matrix can be reconstructed by stacking the vectorized covariance matrices and exploiting the independence of antenna pattern errors among subarrays, where the virtual steering vector is the Khatri-Rao (KR) product of the physical steering vector and its conjugate. By performing a singular value decomposition of this pseudo-signal matrix, the noise-subspace corresponding to the virtual steering vector can be obtained. Thereby, the signal identifiability is improved. Simulation results demonstrate the improvement of the signal identifiability and the superior performance of the E-MUSIC algorithm.