Fall-detection-system

IOT system for fall detection.

The idea and the method

A 3-member team researched the health hazard that accompanies "Falls" and explored different solutions to combat this problem. The team was organised in an agile manner utilising a Gantt Chart and commonly used communication tools, considering it was implemented during covid lockdown.

How it works

For a live demo of the system, look into "DEMO" folder!

Specifications - end result

We wanted to make a system that is:

  • Accurate enough in detecting fall
  • Capable of immediately alerting a 3rd person in call for help
  • Capable of low power, constant use with great autonomy
  • Compact and tiny enough for a comfortable wearable device.
  • Low cost

Tech used:

  • ESP32 DEVKIT1
  • WIFI protocol
  • I2C protocol with interrupts of ADXL345 (accelerometre)
  • MQTT protocol for node control
  • Node-RED for node communication and alerts
  • Alert system both via telegram and via siren node

In the end, such an IOT system was designed, implemented and tested with the following characteristics/functionalities:

fall_detection_system

  • Battery monitoring-low consumption techniques (ex. Deep Sleep mode and interrupts)
  • Fall detection algorithm inside the MCU
  • Wifi connection
  • Wifi credentials storage in Flash memory
  • MQTT protocol and communication with NODE-red server
  • Telegram Chatbot for alerts
  • Siren esp32 node for sound alert.

As an end result, an academic paper was also written, containing all the above mentioned research plus testing results (85% TP 90% TN). The paper will not be published here, but I strongly encourage you to check the DEMO file in this github repo to see the device in live demos!

Authors/Designers

  • Dimitrios Lampros
  • Andreas Chadoumellis
  • Paris Stentoumis

Mentor: Dimitris Karagiannis