基于神经网络的OFDM系统信道估计方法  

Channel Estimation Method for OFDM System Based on Neural Network

作　　者：陈佳勇 徐湛[1] 职如昕 田露 CHEN Jia-yong;XU Zhan;ZHI Ru-xin;TIAN Lu(School of Information and Communication Engineering,Beijing Information Science and Technology University,Beijing 100101,China)

机构地区：[1]北京信息科技大学信息与通信工程学院,北京100101

出　　处：《计算机仿真》2023年第9期350-354,共5页Computer Simulation

基　　金：国家重点研发项目资助(2020YFC1511701);北京市优秀人才资助计划青年拔尖项目资助(2016000026833ZK08)。

摘　　要：为了提高正交频分复用(OFDM)信道估计的准确性,提出了一种将神经网络模型应用于信道估计的方法。在奇偶交错的导频图案下,采用两个神经网络模型分别对奇偶OFDM符号的数据进行训练和估计,将导频处的频率响应作为输入,经过神经网络的输出作为数据符号处的频率响应。仿真结果可以表明,上述方法在信道响应的均方误差性能上比传统的最小二乘(LS)加频域插值的方法有4dB左右的提升。对于系统误码率,以上提出的方法比离散傅里叶变换(DFT)插值也有2dB左右的提升。所提方法是一种复杂度适中且性能良好的信道估计方法。In order to improve the accuracy of orthogonal frequency division multiplexing(OFDM)channel estimation,this paper proposes a method of applying neural network model to channel estimation.Under the odd-even interleaved pilot pattern,two neural network models were used to train and estimate the data of the odd-even OFDM symbols respectively.The frequency response at the pilot was used as input,and the output of the neural network was used as the frequency response at the data symbol.The simulation results show that the method improve the mean square error performance of the channel response by about 4dB compared with the traditional least squares(LS)plus frequency domain interpolation methods.For the system bit error rate,the method proposed in this paper also has an improvement of about 2dB compared with Discrete Fourier transform(DFT)interpolation.This method is a channel estimation method with moderate complexity and good performance.

关 键 词：正交频分复用 信道估计 神经网络 

分 类 号：TN929.53[电子电信—通信与信息系统]