FireSonics is a network of low-cost microphones that listens to the forest and flags a fire within minutes — using sound, which travels around the smoke and terrain that block cameras and satellites.
Hundreds of forest fires ignite every year before anyone notices. Every method we rely on today has the same blind spot: it needs a clear line of sight.
Sound travels around the obstacles that block sight. A fire's crackle carries through trees and even through smoke — which is why FireSonics listens, instead of watching.
Every node runs the same short pipeline, continuously.
A microphone at the node picks up the ambient forest sound.
The audio is sent from the node to a central server.
An AI model checks the sound against the acoustic pattern of fire.
The model classifies the sound as fire, or not fire.
The node's known location is used to estimate where the fire is.
Burning material makes a distinct crackle and pop — but a forest is loud in its own right, and a system that cries wolf loses its usefulness fast. The real engineering problem isn't recognizing a fire recording. It's telling a real fire apart from a forest doing a good impression of one, without missing the real thing.
The model is trained to recognize patterns in audio, rather than a fixed threshold — so it can pick up on the more complex, less obvious features of a fire's sound. To test whether it had actually learned those features, we played it fire recordings it had never heard during training. It correctly picked out a fire sound played from a phone in another room, roughly 9 meters away.
A forest isn't watched by one sensor — it takes many. So each node has to be cheap enough to deploy at scale, without giving up on reliability.
A lower price per unit is what makes a real network possible — but it can't come at the cost of missing the fire it was built to hear.
This year's theme, 'Igniting Innovation', is about forest fires.
In Incheon, we plan to demonstrate the full pipeline, live: nodes placed around a simulated forest, exposed to a mix of real forest sounds, human activity, and a fire. A convincing demo has to do two things at once — catch the fire, and ignore everything else.
FireSonics is the New Technologies Experience project of GPHTech, the robotics team at Gymnázium Pavla Horova in Michalovce, Slovakia — 1st place at FIRST Global Slovakia, and one of 5 finalist teams worldwide.