Toward Real Time Processing of Radar Signals Detecting Elderly People Fall

A novel non-contact method is presented to detect the fall of elderly people in indoor environment. A Continuous Wave (CW) Radar was used to calculate the spectrum of the moving/falling elder using Short Time Fourier Transform (STFT). The Power Burst Curve (PBC) and the acceleration were derived from the STFT and used as features for the classifier. This technique was tested with a 2.45 GHz CW Radar implemented by a Software Defined Radio (Ettus Research USRP N210) and off the shelf low cost components. Especially, the frequency of operation was used because it belongs to free Industrial Scientific and Medical (ISM) band. Furthermore, the frequency was chosen in order to be able to penetrate walls. The antenna selected to be as compact as possible (printed antenna) in order to be placed inside a room avoiding any disturbance. Finally, we present test results classifying between falls or non-falls for both line of sight and behind the wall cases. The comparisons show a good performance in real time applications.