Using the BBC micro:bit and TinyML to sense and respond to the world
Supervisor
Suitable for
Abstract
Co supervised by Johnny Austin Jonny@microbit.orgAt www.microbit.org
"The BBC micro:bit V2 is capable of running TinyML workloads, and there have been some interesting demo applications. However, little work has been done to push the limits of what is possible with the device's built in motion sensor and microphone. This is an open-ended project to look at pushing the envelope of the machine learning tasks possible on the micro:bit. Because the device is widely available and inexpensive, it makes an excellent and accessible platform for datascience and TinyML work. For example, could the micro:bit be used as a wrist-worn device to recognise different activity types, or could it be trained to distinguish bat species, or birdsong? Could we teach it to recognise a range of 'command keywords' for controlling other projects? By demonstrating creative and effective uses for machine learning with the micro:bit you will be feeding into the future feature roadmap of the Micro:bit Educational Foundation as we work out how to teach students around the world about AI and ML with physical computing"
Prerequisites Digital Systems (or equivalent)