Fall Monitoring in the Elderly
Mike Eklund, Roozbeh Jafari, Wenchao Li and Ruzena Bajcsy
The objective in this project is to design, implement, and analyze an architecture which aims at observing the elderly population living alone and sending out an alarm signal if and when some life-threatening situation occurs (for example, the person has fallen, or the person has not moved for more than X number of hours, or some intruder has come into their living space, and so on). The components of this architecture are: wireless accelerometer sensors combined with some signal processing capabilities and a radio to receive and send a signal. These devices will be on the person but also in combination with sensors in the environment. One major problem in signal processing and movement detection in this system is the extensive variations in the morphologies of signals of different patients and under various conditions. We perform data analysis by extracting the extremums. The waveform between every two extremums is considered as a segment. We utilize several metrics on every segment. The metrics include skewness, kurtosis, standard deviation, mean, and the time duration of each segment. We utilize these metrics to perform the classification. The preliminary results show that several basic movements such as standing up, sitting down, lying down, etc., can effortlessly be separated and classified.