Robust Estimation of Angular Power Spectrum in Massive MIMO Under Covariance Estimation Errors: Learning Centers and Scales of Gaussians
Naoto Kaneko, Masahiro Yukawa, Renato L. G. Cavalcante, Lorenzo Miretti · 2024
This paper addresses the problem of the angular power spectrum (APS) estimation in massive multiple-input multiple-output (MIMO) systems. Estimating the APS is crucial, e.g., for improving the efficiency of frequency division duplex systems. Since the problem is an ill-posed inverse problem in general, the proposed method models the APS as a sum of Gaussian functions so that the smoothness of APS is implicitly used as prior information. The weights, scales, and centers are iteratively learned by leveraging the multikernel adaptive filtering framework. Simulation results show the efficacy of the proposed method as well as its robustness against covariance estimation errors.