Performance comparison between GED and GSVD applied to the MUSIC algorithm for interference suppression in multiuser beamforming for mmWave massive MIMO systems

Abstract This article investigates the MUSIC algorithm for detecting the direction of arrival (DOA) of useful signals in the presence of multiple interferers. This algorithm is designed to detect the DOA of useful signals when the powers of the interfering signals are low compared with those of the useful signals. In the presence of strong interference, the MUSIC algorithm is rarely used because its covariance matrix contains both useful and unwanted signals, consequently making it difficult to distinguish between the two. To solve this problem, this article proposes two solutions: replacing the eigenvalue decomposition or singular value decomposition of the received‐signal covariance matrix with the generalized eigenvalue decomposition (GED) or generalized singular value decomposition (GSVD). Both solutions require a reference covariance matrix. The two solutions were compared through simulations for multiuser beamforming applied to massive millimeter‐wave MIMO systems, in the presence of useful and interfering signals, in a fifth‐generation network and beyond. The results demonstrated the effectiveness of both solutions.