A Design Method for Blind Equalizers Using Neural Networks

カテゴリ: 部門大会

論文No: GS8-3

グループ名: 【C】平成14年電気学会電子・情報・システム部門大会講演論文集

発行日: 2002/09/02

タイトル(英語): A Design Method for Blind Equalizers Using Neural Networks

著者名: Nuo Zhang(Chiba University),Xiaoqiu Wang(Chiba University),Jianming Lu(Chiba University),Takashi Yahagi(Chiba University)

著者名(英語): Nuo Zhang(Chiba University),Xiaoqiu Wang(Chiba University),Jianming Lu(Chiba University),Takashi Yahagi(Chiba University)

キーワード: estimation|blind equalization|nonlinear distortion|multilayer perceptrons (MLP)|convergence rate

要約(日本語): This paper considers the problem of blind equalization in digital communication systems by using multilayer perceptrons neural network (MLP). A signal suffers from nonlinear, linear, and additive distortion when transmitted through a channel. Linear equalizers are commonly used in receivers to compensate for linear channel distortion. As an alternative, novel equalizer structures utilizing neural computation have been developed for compensating for nonlinear channel distortion. MLP is a kind of neural network with one or more hidden layers. In this paper, a fully connected three layers MLP is presented for the equalization of QPSK signals in the presence of intersymbol interference, additive white Gaussian noise and nonlinear distortions. The network outputs provide an estimation of the source symbols. The disadvantage of the MLP is the larger complexity. Thus the convergence rate of the MLP based equalizer has been found to be slower compared to the conventional equalizers and to other neural equalizers. This paper also considers the problem that a neural network's convergence rate was slow. The convergence rate became quick by using the proposed method. Simulation results demonstrate the effectiveness of the proposed method compared with Bussgang method.

PDFファイルサイズ: 1,617 Kバイト