Knock Knock Sort #1: Sorting trash by the sound it makes

Hi everyone, and thanks to element14 for picking Knock Knock Sort for My Project Is Trash 2026.

Here is the idea in a few words: a bin that works out what you threw away from the knock it makes when it lands, then drops it into the right bin.

Recycling only works if each item ends up in the right bin. Aluminium can be melted down and reused again and again, and glass and many plastics can become new bottles and containers. That saves raw materials and energy and keeps waste out of landfills and incinerators. But recycling plants receive mixed loads, and every wrong item has a cost. A greasy food container, broken glass in the paper bin or a battery among the plastics can lower the quality of a whole batch, damage the sorting machines or send good material to landfill.

The first sorting step happens at home, in the second it takes to throw something away. The goal of my project is to get that moment right.

Most smart-bin projects use a camera. I wanted to try something different: listen to the item instead of looking at it. A falling item reveals something about its material when it hits the bottom of the bin. Sound and vibrations work in the dark and don't care about labels or dirt (and, for the most paranoid among us, they don't put a camera in your kitchen).

When an item hits a hard surface, it behaves differently depending on the material it is made of. Metal rings, a plastic bottle tends to bounce, paper and cardboard hit softly and stop almost at once.

The basic workflow I have in mind is:

- The IR sensor (CN0569) sees the item when it is thrown in the bin. That tells the system an item is coming, so it is ready to listen at the right moment. The time it takes to pass also gives a hint about the item's size.
- The Nicla Vision's microphone hears the same impact through the air. Sound carries some information that vibration may miss.
- The accelerometer (ADXL372) feels the impact through the surface the item lands on: how hard it hits, which frequencies it excites and how fast the vibration dies away.

A small model running on the Nicla board combines the accelerometer and the microphone data and decides which bin the item belongs in. When it is not sure, the item goes to residual waste, because a wrong item in recycling does more harm than a recyclable item in the residual waste bin.

I expect this solution to have the following advantages over a camera-based smart bin:

1. Works in the dark
2. Ignores labels and dirt
3. Privacy (no images taken)
4. Low data rate fits a small microcontroller, which means low cost and negligible power consumption.

However, it still needs to be proven whether this approach can adapt to how the item lands. That is one of the things I want to investigate during this challenge.