Low-Complexity Semiblind Channel Estimation in Massive MU-MIMO Systems

Abstract

Massive multi-user multiple-input multiple-output (MU-MIMO) systems are a promising solution for achieving high throughput and robust transmission in next generation mobile communications. Achieving the optimal transceiver design in such systems requires an accurate knowledge of the channel state information. However, in massive MU-MIMO systems, the quality of the channel estimates is often degraded by pilot contamination. In this paper, we propose a low-complexity semiblind channel estimation algorithm to mitigate the ill effects of pilot contamination. In the proposed approach, the received signals are first projected onto the subspace with minimal interference, where the bases of this subspace are determined recursively via a low-complexity modified power method. An initial estimate of the projected channel coefficients is then made based on a small number of pilot symbols. Finally, data symbols are detected and the channel estimation is refined alternatively. Compared with existing channel estimation methods, the proposed algorithm has lower complexity due to the subspace projection and innovation process. An asymptotic analysis reveals that the mean square error of the channel estimates is inversely proportional to the length of the data symbols. Simulation results demonstrate that the proposed algorithm outperforms the existing works and alleviates the pilot contamination effects effectively.