This project demonstrates a gesture-based remote control device using the Nordic Thingy:53 device. The development kit connects to a PC through Bluetooth as a HID device. Using gestures you can control media stream or slides of the presentation. Based on accelerometer and gyroscope data, a Neuton machine learning model is created to recognize eight classes of gestures:
- Swipe right
- Swipe left
- Double shake
- Double thumb tap
- Rotation clockwise
- Rotation counter-clockwise
- No gestures (IDLE)
- Unknown gestures
There are two approaches to creating a model for this use-case:
- Using a raw dataset — You can use a raw dataset for model training, which allows the creation of custom models or augmentation of the dataset to train a more robust model.
- Using the provided dataset — You can use the dataset provided. If you wish to use it, go to the Data overview section.