Gamifying exercise allows people to become more motivated and participate more often in physical activities while also being distracted by doing something fun at the same time. This inspired a team of students from the Handong Global University in Pohang, South Korea to come up with a system, dubbed “Move!,” that uses a microcontroller to detect various gestures and perform certain actions in mobile games accordingly.
They started by collecting many different gesture samples from a Nano 33 BLE Sense, which is worn by a person on their wrist. This data was then used to train a TensorFlow Lite model that classifies the gesture and sends it via Bluetooth to the host phone running the app. Currently, the team’s mobile app contains three games that a player can choose from.
There is a dinosaur game that operates similarly to the offline dinosaur game in Google Chrome where the user must jump to avoid incoming obstacles. The jumping jack game alternates between different movements that are mirrored by the player in a certain amount of time. And finally, there is a boxing game where the player punches the air when commanded on-screen.
You can read more about Move! — which was one of the five winning projects in the TensorFlow Lite for Microcontrollers Challenge — here and view/download the code for both the BLE Sense and mobile app on GitHub.