摘要: |
在正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM) 系统中由于快衰落导致信道特征不连续,常规的信道插值方法无法准确反应导频与整个信道之间的关联性。针对这一问题,提出了一种基于宽深超分辨率(Wide Deep Super-resolution,WDSR)网络的信道估计方法,把导频值通过最小二乘估计(Least Squares,LS)初步插值,再通过WDSR网络再次放大重构整个信道的响应。将信道估计插值上采样替换成初步插值和图像超分辨率上采样两步。仿真结果表明,与超分辨率卷积神经网络(Super-resolution Convolutional Neural Network,SRCNN)信道估计算法相比,在不同种类的信道以及导频数下WDSR信道估计方法均方误差性能提升约4.6 dB。 |
关键词: OFDM系统 信道估计 宽深超分辨率(WDSR)网络 超分辨率卷积神经网络(SRCNN) |
DOI:10.20079/j.issn.1001-893x.220817001 |
|
基金项目:国家自然科学基金青年科学基金项目(61701433);云南省科技厅面上项目(2018FB099) |
|
Channel Estimation Based on Wide Deep Super-resolution(WDSR) Network |
XIE Peng,QIAN Rongrong,REN Wenping |
(School of Information Science and Engineering,Yunnan University,Kunming 650500,China) |
Abstract: |
In an orthogonal frequency division multiplexing(OFDM) system,the channel characteristics are discontinuous due to fast fading,and the conventional channel interpolation method cannot accurately reflect the correlation between the pilot and the entire channel.To solve this problem,a channel estimation method based on Wide Deep Super-resolution(WDSR) network is proposed.The pilot value is preliminarily interpolated through Least Squares(LS) estimation,and then the response of the whole channel is amplified and reconstructed again through WDSR network.The channel estimation interpolation upsampling is replaced by two steps of preliminary interpolation and image super-resolution upsampling.The simulation results show that compared with the Super-resolution Convolutional Neural Network(SRCNN) channel estimation algorithm,the mean square error performance of WDSR channel estimation method is improved by about 4.6 dB under different types of channels and pilot frequencies. |
Key words: OFDM system channel estimation wide deep super-resolution(WDSR) network super-resolution convolutional neural network(SRCNN) |