Reiterative Mmse Using Feed-Forward Prior Estimates for Improved Direction Finding
Logan Satterfield, Jonathan W. Owen, Alex Bouvy, Benjamin H. Kirk, Patrick McCormick, Shannon D. Blunt · 2025
Reiterative super-resolution (RISR) is a variant of the reiterative minimum mean squared error (RMMSE) algorithm class, originally developed for adaptive direction finding. RISR was recently experimentally demonstrated to enhance open-air direction-of-arrival estimation while providing robustness to non-ideal calibration errors via practical modeling and incorporation of a gain constraint. RISR is generally initialized with a standard (non-adaptive) beamforming estimate. Here, the recycled estimate (RE)-RISR variant is proposed that instead uses RISR angle estimates from recent snapshots as an initialization, thereby avoiding the need for complete reconvergence at each snapshot. Given sufficient (yet modest) stationarity, RE-RISR significantly reduces the number of required iterations (and therefore computational cost). Simulated performance for a 6 -element uniform circular array (UCA) with stationary and nonstationary incoming signals reveals robustness even in dynamic scenarios.