Files

Please find the workflow in the Workflow-graph.pdf

Device files

main.py - main file to run, measures difference between angle, poll for accelerometer reading
acc.py - accelerometer
mag.py - magnetometer
functions.py - MQTT, additional functions

Server files

mqttServer.py - start up server, accumulate and output results (paho)
mqttServer_publisher.py (not used) - send data from server to device
text.py - API for printing on command line

Correct reading of sensor data

Appropriate sensor settings
Byte data extraction
Two's complement conversion
Magnetometer -> Heading calculation from raw data in x, y and z directions -> Declination angle correction -> Angle conversion to degrees -> Relative angle difference between readings
Accelerometer -> Reading concatenation from high and low registers -> Acceleration magnitude calculation

Sending of data to MQTT broker

Mosquito broker setup on a laptop
Device sends to laptop via MQTT (JSON format)through EEERover network
Server(cloud) receives JSON packets and processes data and outputs statistics of counts and guidance to improve swings

On device processing/formatting

Converting raw data to degrees in 3D
Triggered by push button to switch between swing detection and compass
using accelerometer to set a threshold of detecting swing
Differentiating different types of swing by determining the angle difference using data from magnetometer

Efficient and maintainable code

Files modularised (one for magnetometer, one for accelerometer...)
Only import what is needed from particular modules from machine import ...
Comments where needed

Imaginative product

Teaches person how to play table tennis on his own
Done through measuring swings, with guidance if no swing is detected (E.g. swing left to right more)
Selectable compass function

Addition sensors, other I/O, cloud functionality

Adapted accelerometer onto board for measurement triggering
Only send data when accelerometer above certain magnitude(swing detected), minimising data sent
Server acts like a cloud, processing and output can be sent to app/website/other device, etc

Further improvement

Introduce more sensors for more accurate reading (e.g. gyroscope)
Use machine learning algorithms to analyse an individuals movement, enabling the device to get more accurate as a particular person uses it
Currently strapped on to a glove, perhaps adapting it to a bracelet (e.g. fitbit)/ adapting to be on the bottom of the handle
Utilise algorithms such as the Kalman filter to improve accuracy

Website

link-to-website

Contribution

Website - Octavian Rosu
Everything else - Louis Kueh, Kavin Winson, Un Kei Leong