Kine-Fanatic

Serial output
Raw basketball data
Basketball data aligned with acceleration maxima
ProsE pose

Inspiration

Learn how to effectively throw a ball.

What it does

The Kine-Fanatic allows us to monitor the accelerations of different critical points on the body in movements like throwing, punching, or shooting a basketball. After analyzing multiple actions, we could classify newly recorded actions based on the data we collected.

How we built it

We accomplished data collection and transmission with BBC Micro:bit programmed in C++. Three sensor micro:bits sent acceleration data to the main hub micro:bit which had a serial link to a computer. We then imported the data into Matlab for analysis and classification.

Challenges we ran into

Eliminating gravity from the accelerometer readings.
Transmitting data to the central hub from three active micro:bits.
Finding the same critical points in different recordings.

Accomplishments that we're proud of

Collecting consistent data that was usable for classification.
Collecting all data with a single serial link.
Classifying new recordings as one of three actions.

What we learned

How to filter data based on imperfect sensors.
How to utilize low energy Bluetooth connectivity to interconnect our sensors.
How to use Matlab classification of data.

What's next for Kine-Fanatic

Using this same method on a much larger sample size would have interesting results. Additionally, attempting to classify a movement as "good" and "bad" could lead us to give the user feedback on their action.

Built With

Submitted to

CUES Hackathon 2016 Powered by ARM

Created by

I worked on filtering out gravity in the accelerometer sensor readings. I also worked on the analysis of the collected data in Matlab and devised a way to align data based on the peak acceleration of the movement.

David Bustillos
I worked on the USB serial port connection and the data recording interface. Furthermore, I worked on the data analysis in MATLAB, looking at ways of how to extract the main features, cluster the different motions and make predictions.

Alexander Wettig
I worked on data collection. I explored what data we would be able to make use of (e.g. I also looked at estimating distances using received radio signal strength) and implemented a way for us to collect all data from a single serial link.

Jared Khan