Structured Deterministic Maximum Likelihood Estimation of Polarimetric DOA via Block Coordinate Descent

This letter presents a structure-preserving deterministic maximum likelihood (DML) algorithm for polarimetric direction-of-arrival (DOA) estimation. Retaining each source's waveform-polarization coupling yields a rank-one signal representation and converts the fixed-DOA signal update into a weighted low-rank approximation (WLRA) under an array-manifold-induced metric. Cholesky whitening and singular value decomposition provide a closed-form block update. The resulting block coordinate descent (BCD) algorithm alternates structured signal reconstruction with gradient-based DOA refinement and monotonically decreases the DML objective. Simulations demonstrate accurate and robust estimation with limited snapshots, highly correlated sources, and small angular separations.