A Smartphone Application for a Portable Fall Detection System

Andrew Boehner

Abstract


When a person, especially an elderly person, falls down, it often results in many complications such as life threatening injuries.  In many cases, medical assistance is too late to prevent major complications.  Fall sensing devices are being developed and explored to combat this issue.  However, at the current stage, a common problem is that manual activation of the device maybe impossible due to injuries.  Another problem is that some fall detection systems would require the user to wear uncomfortable equipment.  This could cause people to be reluctant to wear the device.  Therefore, there is indeed a need for portable and intelligent falling detection systems. In this research, we take a novel look at the detection systems.  The aim is a portable, cost efficient, and a user-friendly system that does not interfere with usual habits.  Because people normally carry phones and wear watches daily, we adapt a smart phone that includes medical contacts and a fall detector that is a digital watch, EZ430-Chronos provided by Texas Instruments, to detect the falling and to communicate with emergency contacts.  We study intelligent algorithms to detect falling by using data sensing technologies that includes the 3-axis accelerometer and clock readings, which are recording real-time information from the watch. Our algorithm, embedded in the watch, calculates falling metrics and matches them with falling patterns.  Each movement produces a different degree of acceleration and forms a pattern.  A fall pattern is unique compared to the patterns of other actions.  When there is a matched fall pattern, the watch will send a command through Bluetooth wireless to the smart phone to activate a medical assistance call/message.  Our experiments showed that our system differentiates falling from sitting, walking, running, and most other normal situations.


Keywords


Fall Detection, Smartphone, Digital Watch

Full Text: PDF

Refbacks

  • There are currently no refbacks.