正交时频空(OTFS,orthogonal time frequency space)作为6G关键候选技术之一,能够有效对抗双选择性衰落信道的影响。然而,OTFS系统的信道估计一直是学术界研究的重点和难点。近年来,有研究提出了基于深度学习的OTFS信道估计方案,其运用...正交时频空(OTFS,orthogonal time frequency space)作为6G关键候选技术之一,能够有效对抗双选择性衰落信道的影响。然而,OTFS系统的信道估计一直是学术界研究的重点和难点。近年来,有研究提出了基于深度学习的OTFS信道估计方案,其运用人工智能技术快速捕捉信道变化,但也存在网络规模大、难以满足移动终端轻量化需求的问题。为此,以提高计算效率、降低设备功耗为目标,提出一种基于轻量并行去噪网络的OTFS信道估计算法。该算法结合图像去噪和数据驱动思想,在保留深度学习算法强大的泛化能力的基础上,通过优化网络结构和降低导频功率,降低了移动端的算力成本,为高速移动场景下终端通信的轻量化提供了新的解决方案。所提算法的参数规模仅为现有基于图像去噪的卷积神经网络(DnCNN,denoising convolutional neural network)的15%,大幅降低了网络参数规模和计算复杂度。仿真结果表明,凭借独特的并行结构设计,所提算法弥补了轻量化设计带来的估计性能损失。在五径快时变信道下,所提算法相较于DnCNN实现了4 dB的性能增益。展开更多
Orthogonal frequency division multiplexing （OFDM） systems encounter performance degradations because of the time-varying （TV） channels common in wireless environments. The channel variations within one OFDM symbol...Orthogonal frequency division multiplexing （OFDM） systems encounter performance degradations because of the time-varying （TV） channels common in wireless environments. The channel variations within one OFDM symbol introduce intercarrier interference. In this case, the frequency domain channel matrix is no longer diagonal, hence the corresponding channel estimation is challenging. In this article, two novel TV channel estimation approaches are proposed for the pilot-assisted OFDM systems, where the channel is approximated by the high-order linear model or the piece-wise linear model in time domain. The least square estimation is derived for the two kinds of channel approximations. The simulation is performed based on realistic TV channels with a fairly high Doppler spread. The results show the significant decreasing of the estimation mean square error using the proposed approaches.展开更多
文摘正交时频空(OTFS,orthogonal time frequency space)作为6G关键候选技术之一,能够有效对抗双选择性衰落信道的影响。然而,OTFS系统的信道估计一直是学术界研究的重点和难点。近年来,有研究提出了基于深度学习的OTFS信道估计方案,其运用人工智能技术快速捕捉信道变化,但也存在网络规模大、难以满足移动终端轻量化需求的问题。为此,以提高计算效率、降低设备功耗为目标,提出一种基于轻量并行去噪网络的OTFS信道估计算法。该算法结合图像去噪和数据驱动思想,在保留深度学习算法强大的泛化能力的基础上,通过优化网络结构和降低导频功率,降低了移动端的算力成本,为高速移动场景下终端通信的轻量化提供了新的解决方案。所提算法的参数规模仅为现有基于图像去噪的卷积神经网络(DnCNN,denoising convolutional neural network)的15%,大幅降低了网络参数规模和计算复杂度。仿真结果表明,凭借独特的并行结构设计,所提算法弥补了轻量化设计带来的估计性能损失。在五径快时变信道下,所提算法相较于DnCNN实现了4 dB的性能增益。
文摘Orthogonal frequency division multiplexing （OFDM） systems encounter performance degradations because of the time-varying （TV） channels common in wireless environments. The channel variations within one OFDM symbol introduce intercarrier interference. In this case, the frequency domain channel matrix is no longer diagonal, hence the corresponding channel estimation is challenging. In this article, two novel TV channel estimation approaches are proposed for the pilot-assisted OFDM systems, where the channel is approximated by the high-order linear model or the piece-wise linear model in time domain. The least square estimation is derived for the two kinds of channel approximations. The simulation is performed based on realistic TV channels with a fairly high Doppler spread. The results show the significant decreasing of the estimation mean square error using the proposed approaches.
