<p>In coherent signal direction of arrival (DOA) estimation, traditional decoherence methods often suffer from limited performance because of array aperture shrinkage. While non-circular signals can enhance virtual aperture through covariance matrix expansion, their non-circular phase characteristics can disrupt conventional spatial smoothing algorithms. To address this issue, this paper proposes a novel block diagonal spatial smoothing (BSS) mechanism applied to the extended covariance matrix, which simultaneously enhances the effective array aperture while avoiding the interference of the non-circular phase on the smoothing process. Moreover, since the two-dimensional MUSIC space spectrum requires a two-dimensional search for non-circular phases and DOA, which increases the computational complexity of the subsequent DOA estimation, then the reduced dimensional MUSIC algorithm is employed. Finally, a detailed derivation of the Cramer–Rao bound for non-circular coherent signals is provided. Simulation results demonstrate that the BSS mechanism improves the performance of the classical and state-of-the-art spatial smoothing algorithms compared to circular coherent signals-based algorithms.</p>