Learn

Can your phone detect REM sleep?

Short answer: to a useful degree, yes — and it's improving. A phone can't see brain waves, but sleep stages leave a motion signature: REM comes with near-total muscle paralysis, while other stages carry characteristic movement and stillness patterns a phone's accelerometer can pick up from the mattress. Machine learning can turn that signal into REM estimates — with accuracy that is honest-to-goodness person-dependent, which is why published, per-person results matter.

How sleep is measured in a lab

The gold standard is polysomnography (PSG): EEG electrodes on the scalp, sensors on the eyes and chin, all scored by a technician. It's accurate and completely impractical at home. Every consumer sleep tracker — watch, ring, or app — is an attempt to approximate PSG from cheaper signals.

What motion can tell you

Sleep researchers have used wrist-worn motion sensing (actigraphy) for decades to distinguish sleep from wake. Staging is harder, but each stage has a physical signature:

REM: the body is atonic — nearly paralyzed — while breathing gets faster and more irregular. On a motion sensor: profound stillness with a distinctive respiratory rhythm.
Deep (slow-wave) sleep: still, but with slow, regular breathing.
Light sleep: position shifts, twitches, and movement arousals.

A phone lying on the mattress inherits a version of this signal: its accelerometer picks up body movements and even the mechanical rhythm of breathing transmitted through the bed. The raw signal is noisier than a wrist sensor — which is exactly the kind of problem modern machine learning is good at.

The honest limits

Anyone claiming phone-based sleep staging "just works" for everyone is ahead of the evidence. The real challenges:

Person-dependence: body weight, sleep position, movement habits, and breathing all vary. A model tuned to one sleeper can stumble on another.
The bed matters: mattress firmness, phone placement, and a partner's movement all shape the signal.
Validation is everything: the only meaningful claim is measured accuracy against a reference (PSG in research; a validated wearable as a practical proxy) across many different people.

Why this matters for lucid dreaming

The best-evidenced external trigger for lucidity is a light cue delivered during REM — and the entire history of lucid-dreaming hardware failed on REM timing, not on the cue. If a phone can detect REM windows reliably, the proven trigger finally becomes usable with zero extra hardware.

That's the bet Ludin is testing — an on-device model reading the phone's motion sensors, personalizing to each sleeper over time. Rather than asking anyone to take that on faith, we publish our REM-detection accuracy live, on real people beyond the founding team, as the results come in.

Quick answers

How accurate are phone apps at detecting sleep stages?

It varies widely by app and by person. Motion-based staging can identify REM windows usefully in many sleepers, but accuracy is person-dependent — body type, sleep position, mattress, and phone placement all affect the signal. Look for products that publish validation data.

Does the phone need to be worn or held to track sleep?

No. Placed on the mattress, a phone's accelerometer picks up body movement and the breathing rhythm transmitted through the bed — the same class of signal wrist actigraphy uses, read from a different position.

Why is REM detection important for lucid dreaming?

Because the best-evidenced external lucidity trigger — a light cue — only works when it lands during REM sleep. Imprecise REM timing is the main reason earlier lucid-dreaming devices under-delivered.