This project focuses on recognizing hand gestures and interacting through them with a dashboard. We use the Tunderboard React by Silicon Labs.
We created a uitility to record gestures in /recorder. These recoded data sets are used to train the machine learning algorithm. We chose a Voting classifier with RandomForest and 5-Neighbors as provided by the scikit-learn packages. Dashboard and gesture recognition are implemented in /recognition. Interfacing the Thunderboard React happens through a NodeJS script. For more details on system architecture please refer to /paper where we described the project and its outcome.
Installation on Mac OS:
- Make sure Xcode, NodeJS and Anaconda (Python 2.7) are installed
- Go to
/connector
, runnpm install
- Run
pip install numpy scikit-learn pydot pandas matplotlib
To launch the recording cd into /connector and execute
node gesture.js ../recorder/app.py
To launch the recognition along with the dashboard cd into /connector and execute
node gesture.js ../recognition/app.py
Note that this project was built upon Python 2.7 and relies on Anaconda.