5G高速移动系统中基于BP神经网络的多普勒频偏估计方法

BP neural network based Doppler frequency offset estimation method for 5G high-speed mobile system

提出了一种基于反向传播(backpropagation,BP)算法训练的神经网络的多普勒频偏估计方法。所提方法主要分成线下训练与线上估计两个阶段,首先利用随机多普勒频偏与接收的导频符号构建训练样本,然后利用训练样本对BP神经网络进行线下训练,完成输入与输出数据之间的映射关系,最后基于训练后的网络利用接收导频符号数据,进行线上多普勒频偏估计。仿真结果表明,所提方法的估计性能远远优于现有方法,且具有较低的计算复杂度。

A BP neural network based Doppler frequency offset estimation method was proposed. The proposed method was mainly divided into two stages: offline training and online estimation. Firstly, the training samples were constructed by using random Doppler frequency offset and received pilot signals, and then the training samples were used to train the BP neural network offline, which could complete the mapping relationship between input and output data. Then, based on the trained network, the received pilot signal was used to estimate the Doppler frequency offset online. Simulation results show that the performance of the proposed method is far superior to the existing method, and it has low computational complexity.