Low-complexity Gaussian-Newton method for multi-modulus algorithm-based blind equalization

doi:10.1016/j.sigpro.2022.108722

Return

Low-complexity Gaussian-Newton method for multi-modulus algorithm-based blind equalization

2022-12-01

PRE

DOI：10.1016/j.sigpro.2022.108722

Original

Original request for help

Save

Abstract

En 中文

In this paper, we propose a second-order Gaussian-Newton method for the blind equalization of quadra-ture amplitude modulation (QAM) signals based on the multi-modulus algorithm (MMA). A novel Gaussian-Newton method (GNM) is established to promptly determine the weight vector of the blind equalizer (BE) based on the MMA. Theoretical analysis shows that the proposed GNM can guarantee the BE to iterate along the descent direction because it adopts a positive definite Hessian matrix. Moreover, when the equalizer is close to the optimal one, it is proved that the BE based on the proposed GNM con-verges quadratically similar to Newton methods. More importantly, the Hessian matrix and its inverse matrix are unchanged during the iterative progress, thereby dramatically reducing the computational complexity compared with other Newton-type methods. Simulation results confirm that the proposed BE has slightly better equalization performance and considerably faster convergence speed compared with those of the other Newton-type blind equalization algorithms.(c) 2022 Elsevier B.V. All rights reserved.

Keywords:

Blind equalization

Gaussian-Newton method

Convergence speed

Computational complexity

Convergence stability