Dual-Branch Weighted Root-MUSIC for DOA Estimation in Semi-Passive IRS-Aided Intelligent Wireless Sensing Systems

Abstract

Accurate direction-of-arrival (DOA) estimation is an important spatial-perception capability for intelligent wireless sensing systems, in which communication infrastructure is increasingly expected to support environment sensing and localization. Semi-passive intelligent reflecting surfaces (IRSs) contain passive reflecting elements (PREs) and active sensing elements (ASEs), which provide two complementary spatial structures for DOA estimation. ASE-only methods use only the limited ASE aperture. Weighted ESPRIT (WESPRIT) exploits both ASE and PRE branches, but it extracts rotational factors from the signal subspaces without explicitly exploiting the full noise-subspace polynomial structure, which leaves room for further accuracy improvement. This paper proposes a dualbranch weighted Root-MUSIC method, termed WRoot-MUSIC. Root-MUSIC is applied separately to the ASE sample covariance matrix and the PRE-domain Gram matrix. The known BS-IRS direction is compensated by a fixed complex rotation of the PRE roots, after which the two branches are fused using the eigenvalue-based SNR weights inherited from WESPRIT. Simulations show that WRoot-MUSIC provides lower RMSE than the ASE-only MUSIC baseline in the tested configurations, and lower RMSE than WESPRIT throughout the tested snapshot range and in the tested moderate-to-high transmit-power region. These results demonstrate the potential of WRoot-MUSIC for angular sensing in semi-passive IRS-enabled intelligent wireless sensing systems.