SVD-Fingerprinting-Based Decoupled Channel Estimation for Double-Sided XL-MIMO

Abstract

Channel estimation in double-sided near-field extremely large-scale MIMO (XL-MIMO) systems suffers from prohibitive computational complexity due to intricate transmitter-receiver (Tx-Rx) coupling and exhaustive multidimensional grid searches. To overcome this bottleneck, this letter proposes a low-complexity singular value decomposition (SVD)-fingerprinting based decoupled channel estimation (SFP-DCE) scheme. We reveal that the near-field channel’s singular value spectrum is governed solely by the Tx-Rx coupling factor, strictly invariant to the remaining parameters. By exploiting this property, we introduce a robust offline fingerprinting method to blindly retrieve the coupling factor. This crucial step decouples the near-field channel into a virtual rank-one form, facilitating highly efficient parameter extraction via a two-dimensional discrete fractional Fourier transform (2D-DFrFT). Simulation results demonstrate that the proposed SFP-DCE scheme achieves comparable estimation accuracy while providing a speedup of more than 10 times for massive arrays compared to state-of-the-art baselines.