| 摘要: |
| 在车对万物（Vehicle-to-Everything，V2X）通信中，由于车辆的高速移动，信道呈现快速时变特性。然而，IEEE 802.11p标准中分配的导频数量稀少，传统的信道估计方法无法根据这些有限的导频准确跟踪信道的复杂变化，严重影响了车辆通信的可靠性。为此，提出一种基于图像超分辨率思想的深度学习辅助信道估计方案。该方案将导频处的信道响应视作低分辨率二维图像，采用截断离散傅里叶变换（Truncated Discrete Fourier Transform，T-DFT）插值将低分辨率图像放大至目标尺寸，然后利用基于卷积神经网络（Convolutional Neural Network，CNN）的超分辨率网络模型重建出高分辨率图像，最终实现从有限的导频信息中恢复出整个帧的准确信道响应。仿真结果表明，在信噪比为30 dB时，所提方案的误码率和归一化均方误差分别能够达到1.56×10-7和1.50×10-4。此外，与其他面向车辆通信的深度学习信道估计方案相比，所提方案在较高调制阶数、较大多普勒频移以及较长帧长的复杂环境中性能更优。 |
| 关键词: 车辆通信 信道估计 深度学习 超分辨率重建 卷积神经网络 |
| DOI：10.20079/j.issn.1001-893x.241219003 |
|
| 基金项目:国家自然科学基金资助项目（61801319）；四川省自然科学基金面上项目（2026NSFSC0394）；四川轻化工大学研究生创新基金项目（Y2023285） |
|
| Super-resolution Reconstruction-based Channel Estimation for Vehicular Communications under IEEE 802.11p Standard |
| LYU Chaoluo,LUO Zhongqiang,LEI Yiting |
| (a.School of Automation and Information Engineering;b.Intelligent Perception and Control Key Laboratory of Sichuan Province，Sichuan University of Science & Engineering，Yibin 644000，China) |
| Abstract: |
| In vehicle-to-everything(V2X) communications，the channel exhibits fast time-varying characteristics due to the high-speed movement of vehicles.However，the number of guide frequencies allocated in the IEEE 802.11p standard is sparse，and traditional channel estimation methods cannot accurately track the complex changes of the channel based on these limited guide frequencies，which seriously affects the reliability of vehicular communications.For this reason，a deep learning-assisted channel estimation scheme based on the idea of image super-resolution is proposed.The scheme treats the channel response at the pilot positions as a low-resolution 2D image，and uses truncated discrete Fourier transform(T-DFT) interpolation to enlarge the low-resolution image to the target size.Then，a super-resolution network model based on convolutional neural network(CNN) is applied to reconstruct the high-resolution image，ultimately achieving the recovery of the whole frame’s accurate channel response from the limited pilot information.Simulation results demonstrate that at a signal-to-noise ratio of 30 dB，the proposed scheme achieves a bit error rate of 1.56×10-7 and a normalized mean square error of 1.50×10-4.Moreover，compared with other deep learning-based channel estimation schemes for vehicular communications，the proposed scheme shows superior performance in complex environments with higher modulation orders，larger Doppler shifts，and longer frame lengths. |
| Key words: vehicular communications channel estimation deep learning super-resolution reconstruction convolutional neural network |
《电讯技术》编辑部 