With the growing advancement of wireless communication technologies,WiFi-based human sensing has gained increasing attention as a non-intrusive and device-free solution.Among the available signal types,Channel State I...With the growing advancement of wireless communication technologies,WiFi-based human sensing has gained increasing attention as a non-intrusive and device-free solution.Among the available signal types,Channel State Information(CSI)offers fine-grained temporal,frequency,and spatial insights into multipath propagation,making it a crucial data source for human-centric sensing.Recently,the integration of deep learning has significantly improved the robustness and automation of feature extraction from CSI in complex environments.This paper provides a comprehensive review of deep learning-enhanced human sensing based on CSI.We first outline mainstream CSI acquisition tools and their hardware specifications,then provide a detailed discussion of preprocessing methods such as denoising,time–frequency transformation,data segmentation,and augmentation.Subsequently,we categorize deep learning approaches according to sensing tasks—namely detection,localization,and recognition—and highlight representative models across application scenarios.Finally,we examine key challenges including domain generalization,multi-user interference,and limited data availability,and we propose future research directions involving lightweight model deployment,multimodal data fusion,and semantic-level sensing.展开更多
Massive MIMO is one of tile enabling technologies tbr beyond 4G and 5G systems due to its ability to provide beamforming gain and reduce interference Dual-polarized antenna is widely adopted to accommodate a large num...Massive MIMO is one of tile enabling technologies tbr beyond 4G and 5G systems due to its ability to provide beamforming gain and reduce interference Dual-polarized antenna is widely adopted to accommodate a large number of antenna elements in limited space. However, current CSI（channel state information） feedback schemes developed in LTE for conventional MIMO systems are not efficient enough for massive MIMO systems since the overhead increases almost linearly with the number of antenna. Moreover, the codebook for massive MIMO will be huge and difficult to design with the LTE methodology. This paper proposes a novel CSI feedback scheme named layered Multi-paths Information based CSI Feedback （LMPIF）, which can achieve higher spectrum efficiency for dual-polarized antenna system with low feedback overhead. The MIMO channel is decomposed into long term components （multipath directions and amplitudes） and short term components （multipath phases）. The relationship between the two components and the optimal precoder is derived in closed form. To reduce the overhead, different granularities in feedback time have been applied for the long term components and short term components Link and system level simulation results prove that LMPIF can improve performance considerably with low CSI feedback overhead.展开更多
针对卷积神经网络(convolutional neural network,CNN)计算复杂度高和内存占用大的问题,本文提出了一种基于轻量级CNN的信道状态信息(channel state information,CSI)图像指纹被动定位(LCNNLoc)方法.离线训练阶段,将幅值差矩阵和相位矩...针对卷积神经网络(convolutional neural network,CNN)计算复杂度高和内存占用大的问题,本文提出了一种基于轻量级CNN的信道状态信息(channel state information,CSI)图像指纹被动定位(LCNNLoc)方法.离线训练阶段,将幅值差矩阵和相位矩阵构造成类似于“RGB”的三通道特征图像;同时设计了一个轻量级CNN架构,利用特征图像作为该框架的输入进行训练,在训练结束时将CNN模型保存为指纹数据库.在线定位阶段,采用概率加权质心方法实现了实时的位置估计.实验结果表明,相较于传统方法,LCNNLoc不仅提升了定位精度,还降低了算法运行耗时.展开更多
为应对大规模多输入多输出(Multiple⁃input multiple⁃output,MIMO)系统中信道状态信息(Channel state information,CSI)反馈开销的日益增长,基于深度学习的CSI反馈网络(如Transformer网络)受到了广泛的关注,是一种非常有应用前景的智能...为应对大规模多输入多输出(Multiple⁃input multiple⁃output,MIMO)系统中信道状态信息(Channel state information,CSI)反馈开销的日益增长,基于深度学习的CSI反馈网络(如Transformer网络)受到了广泛的关注,是一种非常有应用前景的智能传输技术。为此,本文提出了一种基于数据聚类的CSI反馈Transformer网络的简化方法,采用基于聚类的近似矩阵乘法(Approximate matrix multiplication,AMM)技术,以降低反馈过程中Transformer网络的计算复杂度。本文主要对Transformer网络的全连接层计算(等效为矩阵乘法),应用乘积量化(Product quantization,PQ)和MADDNESS等简化方法,分析了它们对计算复杂度和系统性能的影响,并针对神经网络数据的特点进行了算法优化。仿真结果表明,在适当的参数调整下,基于MADDNESS方法的CSI反馈网络性能接近精确矩阵乘法方法,同时可大幅降低计算复杂度。展开更多
该文针对闭环多用户MIMO-OFDM系统提出一种基于线性预测的低速率CSI(Channel State Information)反馈方法。根据相关带宽将OFDM子载波划分成多个子带,移动台对每个子带的CSI作线性预测,并对预测误差进行量化编码后反馈给基站;基站使用...该文针对闭环多用户MIMO-OFDM系统提出一种基于线性预测的低速率CSI(Channel State Information)反馈方法。根据相关带宽将OFDM子载波划分成多个子带,移动台对每个子带的CSI作线性预测,并对预测误差进行量化编码后反馈给基站;基站使用相同的线性预测滤波器将反馈来的预测误差恢复成CSI,然后在每个子带上通过迫零-波束赋形实现多用户空间复用。同时,该文还在采用注水定理分配发射功率的条件下,从理论上分析了下行链路信道容量。数值仿真结果显示,每个反馈数据的实部或虚部仅用1bit量化时,本方法仍能够以较高的精度恢复CSI。与目前3GPP LTE标准所采用的基于码书的反馈方案相比,该方法能够在反馈开销相同情况下,有效地抑制同信道干扰,大幅提高系统容量。展开更多
为了满足密集的用户需求,正在发展的蜂窝网络增加了移动系统下的能量消耗,但更广的网络覆盖范围和功耗更低的无线通信系统也给无线通信系统带来了更多的挑战。针对这些持续增长的需求,本文设计了一种能实现能量效率最大化的多小区大规...为了满足密集的用户需求,正在发展的蜂窝网络增加了移动系统下的能量消耗,但更广的网络覆盖范围和功耗更低的无线通信系统也给无线通信系统带来了更多的挑战。针对这些持续增长的需求,本文设计了一种能实现能量效率最大化的多小区大规模多输入多输出(multiple input multiple output,MIMO)系统下行链路的实现方法,提出了在非完美信道状态信息(channel state information,CSI)情况下包含基站天线数、导频复用因子以及用户数量等参数的信干噪比最佳闭式表达,通过最大比合并(maximal ratio combining,MRC)接收技术推导出大规模MIMO系统的下行链路频谱效率,再根据功耗模型得到系统的整体能量效率,利用交替迭代的优化算法进行优化求解,得出最大能效时的相关参数数值。由仿真结果可知,本文所提的多小区大规模MIMO系统的下行链路的实现方法与现有多小区方法相比,能量效率有12.2%的提升,并且对于环境的变化有更好的鲁棒性,对于多小区大规模MIMO系统具有一定参考意义。展开更多
基金supported by National Natural Science Foundation of China(NSFC)under grant U23A20310.
文摘With the growing advancement of wireless communication technologies,WiFi-based human sensing has gained increasing attention as a non-intrusive and device-free solution.Among the available signal types,Channel State Information(CSI)offers fine-grained temporal,frequency,and spatial insights into multipath propagation,making it a crucial data source for human-centric sensing.Recently,the integration of deep learning has significantly improved the robustness and automation of feature extraction from CSI in complex environments.This paper provides a comprehensive review of deep learning-enhanced human sensing based on CSI.We first outline mainstream CSI acquisition tools and their hardware specifications,then provide a detailed discussion of preprocessing methods such as denoising,time–frequency transformation,data segmentation,and augmentation.Subsequently,we categorize deep learning approaches according to sensing tasks—namely detection,localization,and recognition—and highlight representative models across application scenarios.Finally,we examine key challenges including domain generalization,multi-user interference,and limited data availability,and we propose future research directions involving lightweight model deployment,multimodal data fusion,and semantic-level sensing.
基金supported by the National High-Tech R&D Program(863 Program 2015AA01A705)
文摘Massive MIMO is one of tile enabling technologies tbr beyond 4G and 5G systems due to its ability to provide beamforming gain and reduce interference Dual-polarized antenna is widely adopted to accommodate a large number of antenna elements in limited space. However, current CSI（channel state information） feedback schemes developed in LTE for conventional MIMO systems are not efficient enough for massive MIMO systems since the overhead increases almost linearly with the number of antenna. Moreover, the codebook for massive MIMO will be huge and difficult to design with the LTE methodology. This paper proposes a novel CSI feedback scheme named layered Multi-paths Information based CSI Feedback （LMPIF）, which can achieve higher spectrum efficiency for dual-polarized antenna system with low feedback overhead. The MIMO channel is decomposed into long term components （multipath directions and amplitudes） and short term components （multipath phases）. The relationship between the two components and the optimal precoder is derived in closed form. To reduce the overhead, different granularities in feedback time have been applied for the long term components and short term components Link and system level simulation results prove that LMPIF can improve performance considerably with low CSI feedback overhead.
文摘针对卷积神经网络(convolutional neural network,CNN)计算复杂度高和内存占用大的问题,本文提出了一种基于轻量级CNN的信道状态信息(channel state information,CSI)图像指纹被动定位(LCNNLoc)方法.离线训练阶段,将幅值差矩阵和相位矩阵构造成类似于“RGB”的三通道特征图像;同时设计了一个轻量级CNN架构,利用特征图像作为该框架的输入进行训练,在训练结束时将CNN模型保存为指纹数据库.在线定位阶段,采用概率加权质心方法实现了实时的位置估计.实验结果表明,相较于传统方法,LCNNLoc不仅提升了定位精度,还降低了算法运行耗时.
文摘该文针对闭环多用户MIMO-OFDM系统提出一种基于线性预测的低速率CSI(Channel State Information)反馈方法。根据相关带宽将OFDM子载波划分成多个子带,移动台对每个子带的CSI作线性预测,并对预测误差进行量化编码后反馈给基站;基站使用相同的线性预测滤波器将反馈来的预测误差恢复成CSI,然后在每个子带上通过迫零-波束赋形实现多用户空间复用。同时,该文还在采用注水定理分配发射功率的条件下,从理论上分析了下行链路信道容量。数值仿真结果显示,每个反馈数据的实部或虚部仅用1bit量化时,本方法仍能够以较高的精度恢复CSI。与目前3GPP LTE标准所采用的基于码书的反馈方案相比,该方法能够在反馈开销相同情况下,有效地抑制同信道干扰,大幅提高系统容量。
文摘为了满足密集的用户需求,正在发展的蜂窝网络增加了移动系统下的能量消耗,但更广的网络覆盖范围和功耗更低的无线通信系统也给无线通信系统带来了更多的挑战。针对这些持续增长的需求,本文设计了一种能实现能量效率最大化的多小区大规模多输入多输出(multiple input multiple output,MIMO)系统下行链路的实现方法,提出了在非完美信道状态信息(channel state information,CSI)情况下包含基站天线数、导频复用因子以及用户数量等参数的信干噪比最佳闭式表达,通过最大比合并(maximal ratio combining,MRC)接收技术推导出大规模MIMO系统的下行链路频谱效率,再根据功耗模型得到系统的整体能量效率,利用交替迭代的优化算法进行优化求解,得出最大能效时的相关参数数值。由仿真结果可知,本文所提的多小区大规模MIMO系统的下行链路的实现方法与现有多小区方法相比,能量效率有12.2%的提升,并且对于环境的变化有更好的鲁棒性,对于多小区大规模MIMO系统具有一定参考意义。
