Create a low-cost embedded product with camera that is able to learn and recognize a custom and reconfigurable set of 5-10 basic hand signals. For example, this could be used as an alternative method to control digital devices from a far. A mobile application should allow users to 1) configure the device to connect to WiFi/Bluetooth, 2) perform real-time machine learning and training for new hand signals, 3) recognize and effectively identify signals when present. The real constraint of this project is the embedded hardware cost and size, meaning that most computation will need to be forwarded to a cloud server for processing.

CONSTRAINTS
< $5/unit
< 6 months for development and testing
Prediction Accuracy > 75%
False Alarm Rate < 5%
Prediction Speed < 1 second