Power Spectral Analyses to Detect Falls Using 3-D Accelerometers

Abstract

Falls are dangerous for the elderly, often leading to disability and death. While many systems have been developed using accelerometers to detect falls from other daily activities, sensitivity and specificity of these systems decrease when exposed to more rigorous activities such as running and jumping. In this paper, we propose a fall detection algorithm which uses time-frequency analyses of accelerometer data to monitor and detect fall events from normal daily activities. For this study, a wireless sensor system (WSS) is placed at the center of the chest to collect real-time data from subjects performing various daily activities including falling onto a cushioned mat. Following an IRB approved protocol, experiments were conducted on 18 healthy human subjects performing four types of falls (i.e. forward, backward, and left/right sideway falls) as well as normal movements such as standing, walking, stand-to-sit, sit-to-stand, stepping, running and jumping. In total, 324 different activities were collected including 108 fall events. The developed algorithm computes magnitude of 3-axis accelerometer data to check if a critical fall threshold is passed, then calculates the power spectral density of this data for a critical fall duration window (500 ms). Fall events were observed with high energy between 2 and 3 Hz that are distinct from other rigorous activities such as running (3.5–5.5 Hz) and jumping (1–2 Hz). Preliminary results indicate the PSD-based algorithm can detect falls with high sensitivity and specificity greater than 90%, respectively.