Fog

This git repository contains all the code developed and used in the Gait Detection for Parkinson's Disease project.

Prerequisites

To use the python scripts, you will need the following python module(s) installed:

Python 3.6+

See Python's downloads page on downloads and installation.

PyZMQ

PyZMQ is a Python implementation of the ZeroMQ asynchronous messaging library. ZeroMQ allows messages to be passed between two python process (different scripts).

To install on pip3:

pip3 install pyzmq

SMBus

SMBus is a bus protocol that is a subset of I2C protocol. The sensor interface library uses SMBus to communicate with the sensors.

To install on pip3:

pip3 install smbus

bluepy

bluepy is an open source Python library to connect to BLE Peripheral devices such as our Wrist Device. At the moment, this library is limited to connecting to BLE Peripheral devices only.

Link: https://github.com/IanHarvey/bluepy

Other Python Dependencies for Machine Learning

1. joblib

2. scipy

3. pandas

4. xlrd

5. sklearn

   Note: Some dependencies may be missed out here. When running the Python scripts, be sure to look out for errors indicating the lack of dependencies and install accordingly.

Components

DataProvider

This components aims to read values from the 9-DOF IMU (Inertial Measurement Unit) and publishes them on to a ZMQ topic.

However, due to the nature of the circumstances during the project, only data collected from patients are used. DataPublisher.py contains code to read csv files containing these data and publish as if they were from the IMU.

This functionality is handled in the DataPublisher.py script.

Predictor

This component aims to receive values from the 9-DOF IMU and applies the Machine Learning (ML) prediction model. The output of this model will then be published on a ZMQ topic. This functionality is handled in the Predictor.py script.

Feedback

This component will subscribe to the output from Predictor component via the ZMQ topic. It uses this information to provide both the audio and vibratory feedback to the user.

In the case of the audio, the audio is played via the Raspberry Pi OS's audio output stream. The RPi is initially configured to connect to the wireless bone conduction headphone via Bluetooth. When audio is played on the OS's audio stream, the audio will be played on the headphone as well. This functionality is handled in the Feedback.py script.

In the case of the vibratory feedback, there are code written to connect to the Wrist Device's BLE using the bluepy open sourced Python library. The Wrist Device has an onboard BLE module that allows BLE connection. This functionality is handled in the WristFeedback.py script.

Wrist Device

The wrist device contains a Adafruit Flora microcontroller along with a Adafruit BLE module. This allows BLE capability on the Wrist Device. The firmware that runs on that microcontroller can be found in the vib folder. This folder contains a PlatformIO project structure and can be built and uploaded with any software IDE with a PlatformIO extension or just PlatformIO running in a Terminal. See PlatformIO guides for more information.

Other components

There are several other folders containing software as well.

8Feat_Predictor, 9Feat_Predictor and OptFeat_Predictor are other variants of the ML model that were experimented with.

remote_imu is another PlatformIO project for the Adafruit Feather M0 Bluefruit LE microcontroller. Experimental code were written to collect data from the same IMU used in the Central Device and then shared via BLE.

RemoteIMU.py script found in DataProvider folder contains code to connect to remote_imu's microcontroller. This two sets of code are supposed to work in tandem.

config.py is a single Python script that contains all the application parameters for the entire repository. It is supposed to be a centralized collection of these parameters to facilitate the ease in changing configuration.

How to run

With actual hardware (RPi + IMU)

1. First, navigate to DataProvider folder:

   cd path/to/Fog/DataProvider/

2. Then run:

   sudo python3 DataPublisher.py

   You should see the IMU data being read and published.
   Note: sudo may or may not be needed in order to run SMBus on your system.

3. Next on a separate terminal, navigate to Predictor folder:

   cd path/to/Fog/Predictor/

4. Then run:

5. python3 Feature.py

6. Lastly on two other separate terminals, navigate to the Feedback folder.

7. On each of those two terminals, run:

   python3 Feedback.py

   and

   python3 WristFeedback.py

   Note: If you face issues with root privileges try running with a sudo in front of the commands.