Nonlinear blind equalization schemes using complex-valued multilayer feedforward neural networks

Cheolwoo You, Daesik Hong

Research output: Contribution to journalArticle

99 Citations (Scopus)

Abstract

Among the useful blind equalization algorithms, stochastic-gradient iterative equalization schemes are based on minimizing a nonconvex and nonlinear cost function. However, as they use a linear FIR filter with a convex decision region, their residual estimation error is high. In this paper, four non-linear blind equalization schemes that employ a complex-valued multilayer perceptron instead of the linear filter are proposed and their learning algorithms are derived. After the important properties that a suitable complex-valued activation function must possess are discussed, a new complex-valued activation function is developed for the proposed schemes to deal with QAM signals of any constellation sizes. It has been further proven that by the nonlinear transformation of the proposed function, the correlation coefficient between the real and imaginary parts of input data decreases when they are jointly Gaussian random variables. Last, the effectiveness of the proposed schemes is verified in terms of initial convergence speed and MSE in the steady state. In particular, even without carrier phase tracking procedure, the proposed schemes correct an arbitrary phase rotation caused by channel distortion.

Original languageEnglish
Pages (from-to)1442-1455
Number of pages14
JournalIEEE Transactions on Neural Networks
Volume9
Issue number6
DOIs
Publication statusPublished - 1998 Dec 1

Fingerprint

Blind equalization
Feedforward neural networks
Multilayer neural networks
Neural Networks (Computer)
Chemical activation
Quadrature amplitude modulation
FIR filters
Learning
Random variables
Costs and Cost Analysis
Cost functions
Error analysis
Learning algorithms

All Science Journal Classification (ASJC) codes

  • Software
  • Computer Science Applications
  • Computer Networks and Communications
  • Artificial Intelligence

Cite this

@article{7600f17016b14750a94ede78d950d022,
title = "Nonlinear blind equalization schemes using complex-valued multilayer feedforward neural networks",
abstract = "Among the useful blind equalization algorithms, stochastic-gradient iterative equalization schemes are based on minimizing a nonconvex and nonlinear cost function. However, as they use a linear FIR filter with a convex decision region, their residual estimation error is high. In this paper, four non-linear blind equalization schemes that employ a complex-valued multilayer perceptron instead of the linear filter are proposed and their learning algorithms are derived. After the important properties that a suitable complex-valued activation function must possess are discussed, a new complex-valued activation function is developed for the proposed schemes to deal with QAM signals of any constellation sizes. It has been further proven that by the nonlinear transformation of the proposed function, the correlation coefficient between the real and imaginary parts of input data decreases when they are jointly Gaussian random variables. Last, the effectiveness of the proposed schemes is verified in terms of initial convergence speed and MSE in the steady state. In particular, even without carrier phase tracking procedure, the proposed schemes correct an arbitrary phase rotation caused by channel distortion.",
author = "Cheolwoo You and Daesik Hong",
year = "1998",
month = "12",
day = "1",
doi = "10.1109/72.728394",
language = "English",
volume = "9",
pages = "1442--1455",
journal = "IEEE Transactions on Neural Networks and Learning Systems",
issn = "2162-237X",
publisher = "IEEE Computational Intelligence Society",
number = "6",

}

Nonlinear blind equalization schemes using complex-valued multilayer feedforward neural networks. / You, Cheolwoo; Hong, Daesik.

In: IEEE Transactions on Neural Networks, Vol. 9, No. 6, 01.12.1998, p. 1442-1455.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Nonlinear blind equalization schemes using complex-valued multilayer feedforward neural networks

AU - You, Cheolwoo

AU - Hong, Daesik

PY - 1998/12/1

Y1 - 1998/12/1

N2 - Among the useful blind equalization algorithms, stochastic-gradient iterative equalization schemes are based on minimizing a nonconvex and nonlinear cost function. However, as they use a linear FIR filter with a convex decision region, their residual estimation error is high. In this paper, four non-linear blind equalization schemes that employ a complex-valued multilayer perceptron instead of the linear filter are proposed and their learning algorithms are derived. After the important properties that a suitable complex-valued activation function must possess are discussed, a new complex-valued activation function is developed for the proposed schemes to deal with QAM signals of any constellation sizes. It has been further proven that by the nonlinear transformation of the proposed function, the correlation coefficient between the real and imaginary parts of input data decreases when they are jointly Gaussian random variables. Last, the effectiveness of the proposed schemes is verified in terms of initial convergence speed and MSE in the steady state. In particular, even without carrier phase tracking procedure, the proposed schemes correct an arbitrary phase rotation caused by channel distortion.

AB - Among the useful blind equalization algorithms, stochastic-gradient iterative equalization schemes are based on minimizing a nonconvex and nonlinear cost function. However, as they use a linear FIR filter with a convex decision region, their residual estimation error is high. In this paper, four non-linear blind equalization schemes that employ a complex-valued multilayer perceptron instead of the linear filter are proposed and their learning algorithms are derived. After the important properties that a suitable complex-valued activation function must possess are discussed, a new complex-valued activation function is developed for the proposed schemes to deal with QAM signals of any constellation sizes. It has been further proven that by the nonlinear transformation of the proposed function, the correlation coefficient between the real and imaginary parts of input data decreases when they are jointly Gaussian random variables. Last, the effectiveness of the proposed schemes is verified in terms of initial convergence speed and MSE in the steady state. In particular, even without carrier phase tracking procedure, the proposed schemes correct an arbitrary phase rotation caused by channel distortion.

UR - http://www.scopus.com/inward/record.url?scp=0032208430&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032208430&partnerID=8YFLogxK

U2 - 10.1109/72.728394

DO - 10.1109/72.728394

M3 - Article

C2 - 18255822

AN - SCOPUS:0032208430

VL - 9

SP - 1442

EP - 1455

JO - IEEE Transactions on Neural Networks and Learning Systems

JF - IEEE Transactions on Neural Networks and Learning Systems

SN - 2162-237X

IS - 6

ER -