Fall detection with orientation calibration using a single motion sensor

Abstract

Falls are a major threat for senior citizens living independently. Sensor technologies and fall detection algorithms have emerged as a reliable, low-cost solution for this issue. We proposed a sensor orientation calibration algorithm to better address the uncertainty issue faced by fall detection algorithms in real world applications. We conducted controlled experiments of simulated fall events and non-fall activities on student subjects. We evaluated our proposed algorithm using sequence matching based machine learning approaches on five different body positions. The algorithm achieved an F-measure of 90 to 95% in detecting falls. Sensors worn as necklace pendants or in chest pockets performed best.