摘要
针对稀疏的无线多径信道,提出了一种新型的基于深度学习的信道估计算法,即LISTA_SPARSE网络.该方法将传统的迭代收缩阈值算法(ISTA)展开为循环神经网络,搭建带有训练参数的深度学习网络,并在网络中引入稀疏化模块来加强恢复信道的稀疏性.由于目前深度学习的绝大多数体系结构都是基于实值操作和表示的,针对OFDM系统中复数形式的信道估计问题,设计了两种不同形式的LISTA_SPARSE网络:等效实数域的LISTA_SPARSE网络和等效复数域的LISTA_SPARSE网络.仿真结果表明,与现有的LISTA、LAMP等网络相比,LISTA_SPARSE网络有效改善了信道的估计性能,同时,与等效实数域的LISTA_SPARSE网络相比,等效复数域的LISTA_SPARSE网络估计性能提升显著.
Aiming at the sparse wireless multipath channel,a novel channel estimation algorithm based on deep learning is proposed,namely LISTA-SPARSE network.In this method,the traditional Iterative shrinkage thresholding algorithm(ISTA)is expanded into a recurrent neural network,a deep learning network with training parameters is built,and a sparsification module is introduced into the network to strengthen the sparsity of the recovery channel.Since most of the current architectures of deep learning are based on real-valued operations and representations,two different forms of LISTA-SPARSE networks are designed for the complex form of channel estimation in OFDM systems:equivalent real domain LISTA-SPARSE networks and equivalent complex domain LISTA-SPARSE networks.The simulation results show that compared with the existing LISTA,LAMP,etc.,the LISTA-SPARSE network effectively improves the estimation performance of the channel,and the estimated performance of equivalent complex domain LISTA-SPARSE network is significantly improved compared with equivalent real domain LISTA-SPARSE network.
作者
徐微
张慧雪
韩玉莹
王少娜
Xu Wei;Zhang Huixue;Han Yuying;Wang Shaona(School of Electronics and Information Engineering,Tiangong University,Tianjin 300387,China;Tianjin Key Laboratory of Optoelectronic Detection Technology and Systems,Tianjin 300387,China)
出处
《南开大学学报(自然科学版)》
北大核心
2025年第2期87-93,共7页
Journal of Nankai University(Natural Sience)
基金
国家自然科学基金(61901297)。
关键词
深度学习
正交频分复用
稀疏信道估计
deep learning
orthogonal frequency division multiplexing(OFDM)
sparse channel estimation
