A HArdware-Software Framework for High-Reliability People Fall Detection

This paper presents a hardware-software framework for reliable fall detection in the home environment, with particular focus on protection and assistance to older people1. The proposed framework includes three different sensors: a 3D Time-of-Flight range camera, a wearable MEMS accelerometer and a microphone (Fig. 1). These devices are connected with custom interface circuits to a central host embedded PC (e_PC) that collects and processes the information with a multi-threading approach. The information provided by the used 3D camera allows us to describe the environment metrically, in terms of appearance and depth images at QCIF resolution. The wearable three-axis accelerometer, by means of a ZigBee wireless module, delivers acceleration components to the e_PC, which recognizes specific patterns related to falls. Acoustic scene analysis is performed using a commercial Shure microphone. The proposed multi-sensor approach for fall detection has been adopted in order to minimize false alarms