Matrix-Inversion Free Tensor Decomposition for High-Dimensional Channel Estimation in Hybrid Beamforming Wideband Systems for Joint Communication and Sensing
Journal
IEEE Transactions on Vehicular Technology
Start Page
1
End Page
16
ISSN
0018-9545
1939-9359
Date Issued
2026
Author(s)
Abstract
As the antenna arrays are widely adopted in mmWave wireless communication systems, estimation of high-order channel tensor becomes an essential task, especially for joint communication and sensing applications. In this paper, channel estimation of a fifth-order tensor in hybrid-beamforming MIMO-OFDM systems with uniform rectangle antenna arrays is considered. A unified approach, tensor-based orthogonal matching pursuit with rotation (TOMP-R), for estimating azimuth angle of arrival, elevation angle of arrival, azimuth angle of departure, elevation angle of departure, and delay is adopted to achieve gridless estimation. In addition, hierarchical search is employed in all the tensor modes including the delay for complexity reduction. The channel coefficient is also derived by exploiting the property of Dirichlet kernel in the transformed domain without matrix inversion. Residual-incorporated interference cancellation is proposed to effectively mitigate adjacent path interference and capture leaked path energy. From simulation results, the proposed algorithm not only attains good channel estimation results compared to the tensor alternating least squares and structured canonical polyadic decomposition algorithms but also shows precise parameter estimation close to Cramér-Rao bound for sensing.
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Type
journal article