Fall-detection simulator for accelerometers with in-hardware preprocessing

Abstract

Mobile fall-detection systems that use accelerometers (as the ADXL 345) with data pre-processing capabilities, enable processors to remain longer in low power modes and therefore can achieve extended device lifetimes. Since fall-detection on these accelerometers is partially executed in hardware, the development and comparison of fall-detection algorithms requires direct evaluation on the hardware and increases complexity. We introduce a fall-detection simulator for the development and comparison of fall-detection algorithms for accelerometers with and without partial in-hardware pre-processing. In addition comprehensive records of fall-situations and daily living activities were generated for the simulator from recording movements. With the help of the simulator, the sensitivity of a given fall-detection algorithm could be improved from 33% to 93%.