Recently, channel sparsity has been considered as a nature of wireless channels in many researches of intelligent communications, and an increasing number of investigations are conducted by exploiting sparsity of wireless channels, such as deep-learning-based channel estimation and compressive sensing. Growing evidence from channel measurements show that the sparse or approximate sparse distribution assumption of wireless channel is reasonable, however, the observed sparse structure in wireless channels is mostly based on intuitive analysis. Several fundamental aspects of channel sparsity have not been well investigated, and among them, we find that choosing a reasonable measure of channel sparsity has not been fully addressed. To fill the gap, this paper presents several measures for wireless channel sparsity from propagation view and validates them based on realistic channel measurements and data mining methods. Following the spirit that a sparse representation implies a small number of elements contain a large proportion of the energy, the four measures of the number of multipath components (MPCs), channel degrees of freedom (DoF), the Gini index , and the Ricean ${K}$ factor are selected as the potential measures of channel sparsity and fully compared, and the channel diversity measure is used as an indicator of channel sparsity to show the inter-dependency between different measures and channel sparsity. The measurement-based analysis shows that the channel DoF and Gini index provide the best sensitivity and accuracy for measuring channel sparsity, whereas the number of MPCs has the worst performance. Moreover, the widely used channel parameter of Ricean ${K}$ factor is found to have fairly good sensitivity to channel sparsity and can be used for channel sparsity evaluation. The results in the paper can be used to accurately measure sparsity of wireless channel for the design of intelligent communications.

中文翻译：

---

测量无线信道的稀疏性

近年来，在智能通信的许多研究中，信道稀疏已被视为无线信道的本质，并且通过利用无线信道的稀疏性，例如基于深度学习的信道估计和压缩感测，进行了越来越多的研究。来自信道测量的越来越多的证据表明，无线信道的稀疏或近似稀疏分布假设是合理的，但是，在无线信道中观察到的稀疏结构主要是基于直观分析。通道稀疏性的几个基本方面尚未得到很好的研究，其中，我们发现选择合理的通道稀疏性度量尚未得到充分解决。为了填补空白，本文从传播角度介绍了几种用于无线信道稀疏性的措施，并基于实际的信道测量和数据挖掘方法对其进行了验证。秉承稀疏表示意味着少量元素包含大量能量的精神，对多径分量（MPC）数量，信道自由度（DoF），基尼指数 ， 和 赖斯 $ {K} $ 因素选择信道稀疏度的潜在度量并进行充分比较，并使用信道多样性度量作为信道稀疏度的指标，以显示不同度量与信道稀疏度之间的相互依赖性。基于测量的分析表明，信道DoF和Gini指数为测量信道稀疏性提供了最佳的灵敏度和准确性，而MPC的数量则具有最差的性能。而且，广泛使用的Ricean信道参数 $ {K} $ 发现该因子对信道稀疏度具有相当好的敏感性，并且可以用于信道稀疏度评估。本文的结果可用于准确测量无线信道的稀疏性，以进行智能通信设计。