Activity detection using frequency analysis and off-the-shelf devices: Fall detection from accelerometer data

Abstract

Increasingly, applications of technology are being developed to provide care to elderly and vulnerable people living alone. This paper looks at using sensors to monitor a person's wellbeing. The paper attempts to recognise and distinguish falling, sitting and walking activities from accelerometer data. Fast Fourier Transformation (FFT) is used to extract information from collected data. The low-cost accelerometer is part of a Texas Instruments watch. Our experiments focus on lower sampling rates than those used elsewhere in the literature. We show that a sampling rate of 10Hz from a wrist-worn device does not reliably distinguish between a fall and merely sitting down.