Power-Efficient Interrupt-Driven Algorithms for Fall Detection and Classification of Activities of Daily Living

Falls lead to major health problems for the elderly. Immediate help could lower the risk of complications and death and greatly increase the likelihood of returning to independent living. Automatic fall detectors are useful devices that can alert family members and caregivers at those life-critical moments. Traditional accelerometer-based fall studies focus on accuracies and largely neglect the fact that algorithms will mostly be implemented in microcontroller units (MCUs) with limited speed and random access memory. In addition, it is desirable for a fall detector to have a battery life of several weeks or months. This paper presents a fall detection algorithm and a classification algorithm for activities of daily living using a wrist-worn wearable device. Both algorithms are power-efficient and can be implemented easily in an 8-bit MCU. They adopt an interrupt-driven approach based on a modern digital microelectromechanical systems accelerometer which supports interrupts and data buffering. The approach is completely different from conventional algorithms which must examine and process every piece of data sampled at high frequencies. The interrupt-driven approach allows a host MCU to examine significantly less data and only process upon accelerometer or timer interrupts.