Temporal fading is the change in the radio channel between a transmitter and receiver, for example, a fluctuating "number of bars" or signal strength between a laptop and access point, even when neither are moving. Past research in temporal fading treats it only as a problem that degrades wireless communication. Emerging research has shown that temporal fading can be exploited to locate, automatically recognize the activity or gesture, and monitor the health of people in the vicinity of a wireless network. These localization, recognition, and monitoring systems are called RF-based environment monitoring (REM) systems. Improvements in REM technologies could aid in the design of police and search-and-rescue systems that locate breathing people in dangerous or collapsed buildings. As another example, REM technologies deployed in a home could detect falls and detect signs of cognitive or physical decline as part of an aging-in-place sensor system. REM technologies could allow people to diagnose disordered sleeping via wireless devices (e.g., cell phones) left on their bedside. Finally, REM systems could revolutionize indoor and outdoor security systems, helping to protect areas and buildings which are difficult to monitor with existing technologies. To date, no fundamental research in temporal fading mechanisms has been performed to support REM applications. Research in this project considers temporal fading and seeks to establish how it is affected by the movements of people in the environment so that it can be exploited for environmental monitoring.
This project will develop, verify, and exploit new models for the temporal changes that occur to received power and channel response. The scope includes both large-scale motion, i.e., a person walking across a room, and small-scale motion, i.e., a person breathing, or moving an arm. It is known that the effects of motion vary as a function of other multipath fading characteristics of the channel and as a function of the person's position with respect to the transmitter and receiver, but it is not known how. This project will explain these relationships with statistical models that describe temporal fading changes due to human and object motion, to breathing and other periodic motion, as a function of the position of the person or object and a function of other multipath channel characteristics. All models will be verified using real-world measurements, using controlled measurements in an anechoic chamber, and using numerical electromagnetic simulation. From them, the project will develop improved estimators, identify improved features for learning methods, and reduce the manual training required to deploy RF-based environment monitoring systems. A wide variety of device-free localization (DFL) and activity recognition algorithms, as well as gesture recognition and breathing monitoring algorithms, can be improved both in terms of accuracy and efficiency. For example, fingerprint-based DFL methods can be enabled to work as well with fewer training measurements, statistical inversion DFL methods can achieve improved accuracy, and RF-based breathing monitoring systems can be made more robust.
