计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (34): 114-116.

• 网络、通信、安全 • 上一篇    下一篇

基于神经网络的MIMO-OFDM信道估计

才 华,刘广文,陈广秋   

  1. 长春理工大学 电子信息工程学院,长春 130022
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-12-01 发布日期:2011-12-01

MIMO-OFDM channel estimation based on Neural Network

CAI Hua,LIU Guangwen,CHEN Guangqiu   

  1. College of Electronic Information Engineering,Changchun University of Science and technology,Changchun 130012,Chian
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-12-01 Published:2011-12-01

摘要: 提出了一种适用于神经网络框架下的MIMO-OFDM系统的信道估计算法。通过对三层神经网络结构的分析,用两层神经网络实现了一种主成分分析(PCA)最小二乘学习算法。通过导频信息得到MIMO-OFDM信道模型初始值,再用神经网络算法对MIMO-OFDM信道的时变状态参数进行跟踪;采用两层神经网络,由隐层输出对最终输出修正,中间实现可变遗忘因子的改进递推最小二乘学习算法。仿真结果表明,该方法与最小二乘(LS)算法相比,在跟踪时变衰落信道时,估计的均方误差有较大提高,从而有效地改善了接收端的符号检测性。

关键词: MIMO-OFEM技术, 信道估计, 神经网络

Abstract: A channel estimation scheme for a multi-input multi-output orthogonal frequency division multiplexing(MIMO-OFDM) system is presented,which is under the neural network framework.Through the three-layer neural network structure,it achieves a Principal Component Analysis(PCA) least squares learning algorithm.Then,the scheme uses the pilots to estimate the channel impulse response as the initial value,and the neural network algorithm to track the MIMO-OFDM channel time-varying state parameters.To improve the estimation performance,the Neural Network approach uses a variable forgetting factor information to track the varying of the channel.The theoretical analysis and numerical results show that the scheme proposes a high precision and a good performance to track the variations of the fading channels compared to LS algorithm.The simulations show the effectiveness of the new scheme.

Key words: MIMO-OFDM, channel estimation, Neural Network