Real-time Fall Detection with a Smartphone
Soo-Young Hwang, Mun-Ho Ryu, Je-Nam Kim, Yoonseok Yang · Journal of the Korea society of IT services · 2012
In this study, a real-time fall detection system based on a smartphone equipped with three-axis accelerometer and magnetometer was proposed and evaluated. The proposed system provides a service that detects falls in real time, triggers alarm sound, and sends emergency SMS(Short Message Service) if the alarm is not deactivated within a predefined time. When both of the acceleration magnitude and angle displacement of the smartphone attached to waist belt are greater than predefined thresholds, it is detected as a fall. The proposed system was evaluated against activities of daily living(walking, jogging, sitting down, standing up, ascending stairs, and descending stairs) and unintended falls induced by a proprietary pneumatic-powered mattress. With the thresholds of acceleration magnitude 1.7g and angle displacement $80^{\circ}$ , it showed 96.5% accuracy to detect the falls while all the activities of daily living were not detected as fall.