Fall Detection With Wrist-Worn Watch by Observations in Statistics of Acceleration

Abstract

It is common for older people to live alone, which can have tragic consequences if they have an accident and can’t call for help in time. This is particularly acute in an aging society where falling is one of the most common accidents. According to the CDC, 1/4 of people over the age of 65 in the United States fall each year. The development of IoT and MEMS has made it possible to detect falls in time and automatically call for help. The presented fall detection system focuses on the walk-fall-still pattern, collects accelerations through the wrist-worn M5StickC-Plus watch, analyses the data locally in the watch, detects falls using an algorithm based on observations in the statistics of acceleration in one second, and then transmits the alarm signal to a remote healthcare system in real-time via WIFI. The lightweight algorithm has been proven to be 90% accurate in detecting falls, and the system can notify service staff of accidents within 1 second. The features of comfort, lightness, and timeliness make the device more practical than similar products. The low-cost, non-intrusive device can be used in care homes and is also suitable for elderly people living alone.

Keywords

• Fall detection
• acceleration
• wrist-worn
• statistics