An optimum accelerometer configuration and simple algorithm for accurately detecting falls

Abstract

This paper describes the development of an accurate, accelerometer based fall detection system capable of distinguish between Activities of Daily Living (ADL) and fall-events. Using simulated fall-events onto crash mats (under supervised conditions) and ADL performed by elderly subjects, distinguishing between falls and ADL is achieved using an accelerometer-based sensor, mounted on the trunk and thigh of the person. Data analysis was performed using MATLAB® to determine the peak accelerations recorded during eight different types of falls. A fall detection algorithm was proposed using simple thresholding techniques. Results from an evaluation of the detection algorithm show that a fall-event can be distinguished from an ADL with 100% accuracy using a single threshold applied to the resultant acceleration signal from a tri-axial accelerometer located at the chest. Thresholding was thus demonstrated to be capable of discriminating between an ADL and a fall-event, when those falls were simulated falls.