Sensor data fusion for timely emergency alarm

Abstract

Aging of the population is a rising problem for many nations worldwide. What emerges from the trend is an increasing degree of demand for eldercare. Safety of the livingalone elders at home is one of the major concerns. There are in fact commercial services available that utilize an electronic device to enable easy summon of help in an emergency. The state of the art solution infers an unusual inactivity, i.e., auto-detection of an emergency, by a long period of observation on the elder's activity level. This, although avoids false alarms, hinders timely responses to emergencies and jeopardizes the elder's health in a long run. Motivated to enable timely emergency alarm, we propose a multilevel sensor data fusion scheme that infers inactivity of an elder based on accelerometer data collected from multiple locations and time points. The preliminary results indicate that the scheme improves significantly the accuracy of inactivity inference per time point. As a result, the time period required to infer inactive state of the elder is shortened by an order of magnitude.1