Seeing the Unseen: Passive RF Sensing via Learning

这个项目探讨了无线服务和设备的流行，以提取有关环境的有价值的信息。这包括，除其他外，室内环境的占用状态，占用者和运动/运动分类，以及其他可以从无线信号推断的高级信息。获得这种及时和准确的态势感知信息的好处是巨大的。实时入住率信息是智能绿色建筑减少商业和住宅碳足迹的必要条件。准确的运动检测，特别是区分人类和宠物运动的能力，可以帮助提供低成本的家庭安全解决方案。以非侵入性和连续的方式进行自主跌倒检测是为一些最脆弱人群提供长期护理的关键。该项目采用数据驱动的学习方法进行无源射频感知，即通过对现有射频信号的明智处理，提取周围环境的态势感知信息。由于固有的射频损伤、环境波动和收发器位置变化，无源射频传感具有一系列独特的挑战，导致文献中出现了不同的方法。该项目为无源射频传感应用带来了无线通信和射频传播的领域知识和强大的专业知识。这些领域知识对于理解挑战至关重要，并有助于为相关学习问题的制定提供信息，包括监督降维、数据不平衡学习和抽样偏差学习。力求以数据为导向，该项目将建立我们对环境与射频信号接收之间因果关系的认识，并为一些无源射频传感问题提供理论上健全和实际有意义的解决方案。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project explores the prevalence of wireless services and devices to extract valuable information about the ambient environment. This includes, among others, occupancy status of an indoor environment, occupant and movement/motion classification, and other high-level information that can be inferred from wireless signals. The benefits of having access to such timely and accurate situational awareness information are enormous. Real-time occupancy information is essential for intelligent and green building to reduce carbon footprint of commercial and residential buildings. Accurate motion detection, and in particular, the ability to distinguish motions between human and pets can help provide low-cost home security solutions. Autonomous fall detection in a non-intrusive and continuous manner is key to providing long-term care for the well-being of some of the most vulnerable populations.This project takes a data-driven learning approach for passive RF sensing, i.e., extracting situational awareness information of the ambient environment through judicious processing of existing radio frequency (RF) signals. Passive RF sensing has a unique set of challenges due to inherent RF impairments, environment fluctuation, and transceiver location changes, leading to divergent approaches in the literature. This project brings domain knowledge and strong expertise in wireless communication and RF propagation to passive RF sensing applications. Such domain knowledge is critical for understanding the challenges and helps inform the formulation of associated learning problems including supervised dimensionality reduction, learning with data imbalance, and learning with sampling bias. Striving for a data-driven approach, the project will build up our knowledge of the cause-and-effect relationship between environment and RF signal reception and provide theoretically sound and practically meaningful solutions to a number of passive RF sensing problems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.