iFall: The Unexpected-Fall Detection Device

A Complex Air Pressure Sensor The iFall System Typical Data The Architecture

Keywords: Ambient Assistant Living (AAL), Unexpected Falls, Acceleration Sensors, Air Presure Sensors, Timely Fall Detection, Fall Prevention, Femoral Neck Fractures

Motivation: Above the age of 65, More than fifty percent of the elderly people of 65 years in age or older experience an unexpected fall once per year. The problem is that many of these unexpected falls go along with major injuries, primarily femoral neck fractures. With about 160,000 cases per year, these unexpected falls induce about two billion euros per year to the German health care system. In addition, every case generates further costs and inconveniences to the subject and his or her social environment. For these reasons, the prevention of unexpected falls and/or their effects have received recent attention. However, the currently available protection devices either do not receive an acceptable acceptance rate and/or are activated only after the fall.

Goals: The major goals of the project are the development of a small embedded system that has the following properties:
(1) small, light, and non-intrusively wearable,
(2) using a protector device that can be seamlessly integrated into regular clothes,
(3) timely activation of a protector, at least 300 s before touch down, and
(4) high reliability and a very low chance of false alarms.

Approaches: This project has explored the following approaches:
(1) employment small embedded systems with only low computational resources and low energy consumption,
(2) integration of both air pressure and acceleration sensors, and
(3) development of intelligent, biologically inspired algorithms.

Results: The first prototype is based on the motion sensor board (MoSeBo) that was developed at FHG-IGD. Furthermore, the prototype uses a new class of fall detection algorithms that correlated subsequent acceleration sensor values as well as the dynamic change of an air pressure sensor. In a larger number of dedicated (artificial) fall experiments (done in a gym), the device has detected the fall at least 300 ms before touch down.

A Performance Figure

Research Team: Dipl.-Ing. Gerald Bieber, Marian Lüder, and Ralf Salomon


Selected Publications: