Low power embedded system for pre-fall detection application
Neeraj Rathi, Ashok Kumar Thella, Monika Kakani, Maher Rizkalla · 2017
Fall has become a serious medical problem, and sometimes leading to physical disabilities and death. This led researchers to pursue automatic monitoring systems for detecting the falls before they occur. Much of the existing efforts have successfully achieved a hardware system which provides a fall pattern after and prior to the fall. However, the existing fall detection systems are still deficient in achieving power optimization and optimum sensor structures. In this study, we have demonstrated a pre-fall detection system using accelerometers and gyroscopes, associated with EFM32GG - ARM Cortex -M3 based 32-bit microcontroller and MicroSD. The system is designed to get optimal low power consumption, better sensitivity & specificity, and early pre-fall detection, to trigger safety devices. The algorithm enables the system on interrupts, which then calculates angular positons of the subject using motion sensors and combines the result with Signal Vector Magnitude (SVM) and Signal Magnitude Area (SMA) to detect the pre-fall accurately. The system was constructed and tested. The practical model consumes 409.1μA in deep sleep mode and 15mA in the active mode with 92% sensitivity and 98.07% specificity. The system stays in sleeping mode during majority of user activity of daily living (walking, sitting), thus improving the battery life of the device. The system also successfully differentiates between the activity of daily living (ADL) like walking, sitting, running, and climbing stairs with actual fall. This paper details the power optimization and consumption, the sensor system architecture, the software algorithm for pre-fall detection and the low power hardware wireless.