Breathing Turns into Sleep ✎ express

Original: "Physiology as Language: Translating Respiration to Sleep EEG"
arXiv:2602.00526 · 2026-01-31 · arXiv non-exclusive · independent analysis · 1 min
A neural network learns to draw the brain's electrical waves from nothing but breathing.

At night, a person breathes, and the rhythm of that breathing can tell you a lot about their sleep. But usually, doctors look at the brain's electrical waves — recorded by sensors on the head. New work shows how to reconstruct those waves from breathing alone, without touching the head.

The model acts as a translator from the language of the body to the language of the brain. It takes a breathing recording and turns it into a sequence of simple pieces, just as a musician breaks a melody into notes. Then, from those notes, a full picture of the brain's electrical activity is assembled. The model was trained on recordings from more than 28,000 people, and the reconstruction error was only 7%.

From the reconstructed waves, you can determine a person's age with an accuracy of 5 years, sex with a probability of 81%, and sleep stage with an accuracy of 84%. Almost as good as from real recordings.

The most interesting part is that the model works even with contactless sensors. Radio waves bounce off the body and pick up breathing, and then the model turns those reflections into a picture of sleep. This paves the way for quiet diagnostics without wires or glued-on electrodes.

The error in estimating age from synthesized brain waves is just 5 years — comparable to the accuracy achieved with real recordings.

Scientists
Jascha Sohl-Dickstein
Concepts
electroencephalography EEG seizure detection electrocardiogram ai speech analysis speech recognition neural language acquisition voice assistant gene expression synthetic data