Can a Sensor Across the Room Really Tell If You're Asleep?

The Honest Answer Is "Some Of It"

You're lying still in a dark room. A small device a few feet away, with no camera and nothing touching your skin, claims to know whether you're asleep. That should sound a little suspicious.

We build one of these things, so we have an obvious bias. We're going to try to be useful anyway, because the marketing in this category is bad enough that a plain explanation is worth more than another set of claims.

Here's the short version. Radio sensing is very good at knowing whether a body is there, roughly where it is, and how much it's moving. It gets shakier the moment your bedroom stops looking like a research lab. And there are things it simply cannot see, no matter how good the software gets.

Key Facts

• A 2024 preprint (not yet peer reviewed) compared radio-based sleep staging against full sleep studies in 849 people and reported 81% agreement on 30-second chunks (arXiv preprint)
• Accuracy depends heavily on distance and obstacles: results are "highly dependent on the distance between the device and the patient, as well as on the presence of possible obstacles in the path" (Sensors, 2023)
• Shared beds are a known weak spot. Contactless systems "may struggle to differentiate between two occupants" (Sensors, 2025)
• One research radar setup discarded every reflection closer than 40 cm and farther than 3 m just to get clean data (Sensors, 2025)
• A 2025 meta-analysis of radar apnea detection pooled 20 studies and 1,540 people, with individual studies ranging from 5 to 196 participants (Diagnostics, 2025)
• In a seven-device head-to-head against a sleep lab, every stage-tracking device — including contactless ones — differed from the lab on light sleep (Sleep, 2021)

How A Radio Wave Notices A Sleeping Person

The idea is simpler than it sounds. The device sends out a radio signal. The signal bounces off everything in the room — walls, furniture, your body — and comes back. The device listens to what returns.

Two useful things fall out of that echo.

Distance. By timing how long the echo takes, the device can sort reflections into distance bins. Something at 1.2 meters is a different return than something at 2.5 meters. That's how it separates "the person on the bed" from "the dresser."

Change. This is the important one. If a reflection at a fixed distance is shifting back and forth by a tiny amount, the signal that comes back is subtly altered. Your body is never truly still. Your chest rises. Your shoulder settles. You roll. Each of those produces a different signature in the returning signal.

So the device isn't seeing you. It's watching how the shape of the room's radio echo changes over time, and inferring a body from it. If you want the engineering detail — chirps, range-Doppler, antenna arrays — we wrote that up in the 60GHz radar deep dive.

Critically, none of this produces an image. There's no picture to leak. That's a real privacy property, not a slogan, and it's why we built around it.

What It's Actually Good At

The strong suits are the coarse, physical facts.

Presence. Is somebody in the bed? This is close to trivial for radio sensing. A human body is a big, warm, moving reflector.

Bed exit. Did the body leave? Also easy, and useful — bed exits at 3am, counted over weeks, tell you something a single lab night can't.

Gross movement. Rolling over, kicking, restlessness, the long stretches of near-stillness. Movement is exactly the thing this sensing modality is built to catch.

Position. Which way you're facing. Harder than presence, easier than sleep staging, and one of the most underrated things you can log across many nights.

Timing. Roughly when you settled, roughly when you got up. Not to the second. To the useful minute.

The lab work backs the ceiling being higher than that. In a 2024 preprint — again, not peer reviewed — researchers matched a radio-based system against full clinical sleep studies in 849 people and reported 81% agreement on 30-second epochs across four categories: wake, light, deep, and REM [^1]. That's a real number from a real comparison. It's also a number produced by a research team, on their own hardware, in a study population, which is not the same thing as your bedroom.

Where It Gets Hard

This is the section the product pages skip.

Two people in one bed

This is the big one. If two bodies are in the sensing field, their reflections mix. A 2025 review of home sleep monitoring technology puts it plainly: shared beds "introduce measurement errors," and these systems "may struggle to differentiate between two occupants" ¹.

Separating them is an active engineering problem, not a solved one. Anyone who tells you their contactless device handles a shared bed perfectly is describing a research goal.

Distance and geometry

Radio signals get weaker with distance, and the returning echo gets weaker faster than the outgoing signal does. That systematic review of contactless cardiac and respiratory sensing found accuracy is "highly dependent on the distance between the device and the patient, as well as on the presence of possible obstacles in the path" ².

Real research setups deal with this by constraining everything. One 2025 radar study mounted the sensor a meter above the mattress, told participants which way to point their feet, and then threw out every reflection closer than 40 cm (antenna crosstalk) or farther than 3 m (signal bouncing off walls and arriving late) — and still only had six participants ³.

Your bedroom does not come with those constraints.

Obstacles and clutter

Thick blankets, a pile of laundry, a headboard, a body angled behind a mound of pillows. Every one of those changes the path the signal takes. The same review lists "background interference, motion artefacts, and electromagnetic interference" alongside "body orientation" and "changes in sleep posture" as ongoing challenges ².

Anything else that moves

A radio sensor doesn't know what a dog is. It knows there's a moving reflector on the bed. A cat that settles on your legs, a partner reaching for water, a ceiling fan — all of it lands in the same signal the algorithm is trying to read a human out of. Good systems filter aggressively. No system filters perfectly.

What It Fundamentally Cannot See

Some limits aren't engineering problems. They're category problems.

A sleep lab measures brain electrical activity, eye movement, and chin muscle tone. Those are what define a sleep stage. Radio sensing has access to none of them. Everything a contactless system says about "deep sleep" is inference from body motion and timing, not measurement of your brain.

That gap shows up in the data. In a seven-device comparison against clinical sleep studies — a test that included three non-wearable, contactless devices — every stage-tracking device differed from the lab on light sleep, and several underestimated REM ⁴. We've written more about why sleep-stage numbers deserve skepticism.

The Lab-To-Bedroom Gap

Here's the thing that should make you cautious about every number in this category, ours included.

A 2025 systematic review of radar-based apnea detection pooled 20 studies with 1,540 participants. Individual studies ranged from 5 to 196 people. The authors flagged substantial variation between studies and noted that most validation happened in controlled environments rather than homes ⁵.

That's the pattern across contactless sensing research. Known geometry. Instructed sleeping. Single occupant. Clean room. Short duration. Then the same technology ships into a bedroom with a partner, a dog, a fan, a duvet, and a bed pushed against a wall.

The research is real, and the gap is also real. Both things are true at once.

Where Komori Fits

Komori is being built as a two-piece system. A small sensor mounts above the bed — that's where the sensing happens — paired with a base unit on the nightstand that handles power, Wi-Fi, and local processing. No camera. Nothing worn.

Mounting above the bed is a deliberate answer to the geometry problem in this post. A sensor looking down at the sleeping surface has a shorter, cleaner path to the body than one looking across a room full of furniture.

When it ships, Komori is designed to log sleep position, movement and restlessness, bed-exit events, room conditions like temperature, humidity, CO2, light and noise, and sound events. That's the list. It's a general wellness device — it doesn't diagnose anything and it isn't a sleep study.

Notice what's not on that list. We're not building it to report breathing rate, heart rate, or blood oxygen. We're not claiming to separate two people in a shared bed. We keep a running public list of what Komori can't do for exactly this reason.

The Bottom Line

A sensor across the room can tell you a lot about a body: that it's there, where it is, how much it's moving, and when it left. That's useful information when you collect it for months instead of one night.

It cannot read your brain. It cannot tell your dog from your knee with perfect reliability. And it cannot diagnose you.

If something about your sleep worries you — snoring, gasping, exhaustion that doesn't lift — no contactless device is the right next step. A sleep specialist is.

---

Footnotes

1. Mathunjwa BM, et al. "A Comprehensive Review of Home Sleep Monitoring Technologies: Smartphone Apps, Smartwatches, and Smart Mattresses." Sensors, 2025. ↩

2. Boiko A, Martínez Madrid N, Seepold R. "Contactless Technologies, Sensors, and Systems for Cardiac and Respiratory Measurement during Sleep: A Systematic Review." Sensors, 2023;23(11):5038. ↩ ↩²

3. Szmola B, et al. "Feasibility of Radar Vital Sign Monitoring Using Multiple Range Bin Selection." Sensors, 2025. ↩

4. Chinoy ED, et al. "Performance of seven consumer sleep-tracking devices compared with polysomnography." Sleep, 2021. ↩

5. Tran NBMH, et al. "Radar-Based Detection of Obstructive Sleep Apnea: A Systematic Review and Network Meta-Analysis of Diagnostic Accuracy Across Frequency Bands." Diagnostics, 2025. ↩