Automatic in-door fall detection based on microwave radar measurements

The use of a Continuous Wave (CW) Doppler radar is proposed for non-invasive automatic detection of human falls. This radar technology can be used since fall incidents can be characterized by changes in speed. In this paper we show that speed measurements obtained from different activities, using a radar fixed on the ceiling, can automatically discriminate between fall incidents and other activities with good accuracy. The activities we consider are falling, walking, running, and sitting. Off-the-shelf machine learning techniques are used to estimate an activity classification model.