Logo image
Enhanced Signal Processing with Quaternion-Valued MIMO: Theory, Design and Applications
Journal article   Peer reviewed

Enhanced Signal Processing with Quaternion-Valued MIMO: Theory, Design and Applications

Jing Zhu, De Mi, Pei Xiao, Rahim Tafazolli and Fan Wang
IEEE Transactions on Communications, Vol.In Press(In Press)
05/11/2026

Abstract

Quaternion-valued signal processing Multipleinput multiple-output (MIMO) Frobenius inversion Rank adaptation Multi-user MIMO (MU-MIMO)
The advancement of wireless communication toward the sixth-generation (6G) necessitates sophisticated signal processing techniques to accommodate high-dimensional data, massive connectivity, and efficient spectrum utilization. Quaternion-valued signal processing has emerged as a compelling approach, leveraging the four-dimensional structure of quaternion algebra to achieve compact signal representation and enhanced system performance. In this paper, we investigate the theoretical foundations and signal-processing design of quaternion-valued multiple-input multiple-output (QV-MIMO) systems and clarify their relationship with conventional complex-valued MIMO (CV-MIMO). Under Gaussian channel and signaling assumptions, we establish the statistical-matching conditions between the QV and CV representations and identify when exact conditional equality in distribution can be achieved. We further distinguish this statistical relationship from the receiver-dependent mechanisms governing BER performance. In the 2 × 2 case, the quater-nion representation enables scalar equalization, whose post-equalization noise is isotropically scaled, whereas conventional CV-ZF detection is subject to singular-direction-dependent noise enhancement. We demonstrate this through BER comparisons and separate the effect of the quaternion representation from that of the effective equalization structure. For higher-dimensional systems, we employ a Frobenius inversion (FI)-based detector, which retains the cubic complexity of direct CV inversion but reduces the leading-order arithmetic cost by 46.875% for the matched system dimensions. We further extend QV-MIMO to multiuser uplink detection and downlink precoding with rank adaptation. In addition, the 2×2 QV receiver is shown to achieve the same post-detection SNR as the Alamouti scheme under a static channel while operating within a single signaling interval. Simulation results under uncoded, imperfect-CSI, mobility, and LDPC-coded settings showcase the performance and computational benefits of the proposed quaternion-domain processing framework under the considered system models. Index Terms—Quaternion-valued signal processing, Multiple-input multiple-output (MIMO), Frobenius inversion, Rank adaptation , Multiuser MIMO (MU-MIMO).
pdf
QV-MIMO_final3.97 MB
Author's Accepted Manuscript Restricted. Access maybe granted on request., This file will be open access upon publication. CC BY V4.0
url
https://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=26View
Journal website

Metrics

1 Record Views

Details

Logo image

Usage Policy