Human Fall Detection Using Distributed Monostatic UWB Radars

Abstract

In a rapidly growing population of elderly people, a system that will allow more of them to stay home for a longer time is of large value to society. A large factor in elderly needing to move from their homes and into caring homes are the risk of injuries from falling. A system that incorporates fall detection will reduce the time it takes for aid to come will therefore be desired. It is also desirable for a system where the person under observation are not required to use any type of sensory equipment. This thesis explore the use of distributed radars for the purpose of fall detection without wearable sensors.In this thesis it will be presented methods for range estimating using returned reflections from the environment with the use of Ultra Wideband Radars and how to use these estimates for target localization. Through the use of weighted least square and nonlinear least square method, positions of objects are estimated in three dimensions. The problems of bias in the range estimates from the body not being a point are analyzed and the relation between radar placement and object position are also discussed. Lastly, the thesis presents a method for fall detection through determinate whether a person is lying on the floor over a time period.