AI can turn a cloud of scuba bubbles into an early warning for divers

University of Minnesota researchers built an underwater AI robot that estimates a diver’s breathing rate by analyzing exhaled bubble patterns.

Using ordinary camera footage, the system counts bubble bursts and classifies breathing as below, normal, or above range.

It was tested in Minnesota lakes and the Caribbean, and could provide noninvasive monitoring without extra wearable sensors.

The approach may help flag unusual breathing as an early warning sign, but it has not yet been proven to reliably detect stress or emergencies.