Although a large percentage of our trash can be recycled, only a small percentage actually makes it to the proper facility due, in part, to being improperly sorted. So as an effort to help keep more of our trash out of landfills without the need for extra work, Samuel Alexander built a smart recycling bin that relies on machine learning to automatically classify the waste by the sound of collision and sort it into separate internal compartments.

Because the bin must know what trash is being tossed in, Alexander began this project by first constructing a minimal rig with an Arduino Nano 33 BLE Sense to capture audio and send them to an Edge Impulse project. From here, the samples were split into 60 one-second samples for each rubbish type, including cans, paper, bottles, and random background noise. The model, once trained, was then deployed to the Nano as a custom Arduino library.

With the board now able to distinguish what type of garbage has been thrown away, Alexander got to work on the remaining portions of the smart bin. The base received a stepper motor which spins the four compartments to line up with a servo-actuated trap door while a LiPo battery pack provides power to everything for fully wireless operation.

To read more about how this bin was created, you can visit Alexander’s write-up here on Hackaday.io.