Fall Sensor

Presentation | Hardware Schematic | User Manual | Sprint Summary

Embedded prototype of fall detection system based on incremental development model and motion data analysis app.

The project constitutes the appendix to my Engineering Thesis written for AGH University of Science and Technology in academic year 2017/18.

Abstract

• whole project based on incremental development model with one-week sprints
• several releases from which any could play a role of MVP
• provided hardware and software prototype

Development steps

1. Build distributed measurement system for collecting data from the sensor and process it using PC.
2. Perform ADL (activity daily living) scenario and simulate accidents.
3. Implement fall detection algorithm.
4. Add emergency system.

System overview

Requirements

Software

• ARM mbed SDK (release 150) and online IDE
• LabVIEW 2016

Hardware

• NUCLEO-L476RG, ST Microelectronics
• 10 DoF IMU Sensor, Waveshare
• USB-UART converter
• microSD card slot
• buzzer, LED, button
• passive elements
• 4x AA batteries or other supply

How to use

Developer

1. Prepare a prototype, based on schematic.
2. Import project to online compiler. You have to do it manually - sorry!
3. Import SDFileSystem as well as mbed (watch out - some newer versions could be incompatible).
4. Make sure the pinout meet your expectations.
5. Program the device.

End User

The device works in two modes:

• fall detection
• data acquisition

To make use of its features go through the User Manual, section: A.2. How to use.

Communication

Serial

There are couple of commands for serial communication available:

Packet frames

Two links can be distinguished.

MCU-SENSOR

Medium: I2c | Payload: 14 B | Acceleration, rotation 3-axis 16b and 16b temperature

SENSOR-DATA ANALYSIS APP, SENSOR-STORAGE

Medium: UART/SPI | Payload: 51-24027 B | Configuration, measurement samples

Configuration

Device has an ability to be configured with parameters shown below.

{
  "SensorConfiguration": {
    "AccelDLPFFrequency": "5HZ",
    "AccelFSRange": "8G",
    "Frequency": "100HZ",
    "GyroDLPFFrequency": "5HZ",
    "GyroFSRange": "1000DPS",
    "InterruptMode": "DATA_RDY",
    "InterruptPinMode": "OPEN_DRAIN_FALLING_EDGE",
    "Resolution": 16
  }
}

The configuration file is stored in the external memory. Can be changed via:

• serial port (as described earlier)
• data analysis app

Readouts

Are stored as raw packets in external memory and sent via serial port on demand. For acquisition, circular buffer has been implemented. 10-second buffer for data is applied.

Fall detection

Block diagram

Always 5 s before and 5 s after event are available thanks to double-buffering.

Data acquisition

Block diagram

Up to 150 000 can theoretically be saved in 4 GB external flash. Note that a data transfer between device and PC isn't efficient enough for sensible transfer of such a big amount of data.

Motion data analysis

Responsibilities | File extension: *.fsdat

Can be made with the app mentioned earlier. User interface is simple but readable. Features:

• browse and comment readouts
• configure device
• export data to CSV/JSON
• change graph display mode

Registered fall (in data acquisition mode)