Simulation of IoT-oriented Fall Detection Systems Architectures for In-home Patients

Abstract

Fall detection (FD) systems enable rapid detection and intervention for people who experience falls, a leading threat to the elderlys health and autonomy. Most of these systems conform to an IoT reference architecture which may include multiple sensing mechanisms to balance the advantages and drawbacks of each alternative. However, developing such a heterogeneous system may be costly and quite resource and time-demanding. This paper presents a Discrete Event System Specification (DEVS) simulation model for FD systems that compares the accuracy of nine different systems architectures that combine traditional wearable and non-wearable sensing devices in the acquisition layer. We perform simulations for each architectural arrangement using four public datasets of FD systems, totaling 36 simulations. Results reveal that an FD accuracy of 96.67% is possible with an investment of almost 6,000 US. Besides, spending 36 times less (around 150 US), designers and clients could acquire an FD system composed of wearable and non-wearable devices with an accuracy of 91%, i.e., only 5% less than the most expensive alternative.