How Accurate Are Radar Sleep Trackers? Breathing, Heart Rate and Sleep Stages

Radar sleep trackers look confident in an app. The papers are narrower. Breathing is the number that holds up. Heart rate is less often measured by radar. Stages are the soft part.

Disclosure: I am Dovy, founder of Komori Care. We are building a contactless bedside sleep monitor, coming 2027, wellness only. There is no Komori accuracy number to quote.

Quick answer

• Breathing: Somnofy measured respiratory rate with 0.18 breaths a minute mean absolute error against a respiration belt, with 84% coverage (JMIR Biomedical Engineering 2022). Accuracy dropped during REM and wake.
• In older adults, Somnofy's nightly breathing-rate error against PSG was 4.64% MAPE, and the radar in that study measured breathing, not heart rate (PMC11387924).
• Sleep stages: Sleepal's vendor preprint reports 92.8% sleep/wake and 78.5% four-stage accuracy on 1,022 recordings in the Normal group (arXiv 2604.16442). Overall four-stage accuracy in that preprint is 77.2%. Consumer devices do not replace a sleep study (Chinoy et al., SLEEP 2021).

For the mat vs nightstand split, see under-mattress vs bedside radar. For the lamp price, see whether the Sleepal AI Lamp is worth it.

Breathing is the strong radar number

The 2022 Somnofy paper used a respiration belt as the reference in healthy adults. Coverage was 84% of time in bed. Mean absolute error was 0.18 respirations per minute. Deep sleep was the easiest (MAE 0.11, coverage 98%). REM and wake were worse (MAE 0.30 and 0.33, coverage 75% and 47%).

The aging lab (35 adults, 65 to 83) compared Somnofy to PSG-derived breathing. Nightly MAPE was 4.64%. Somnofy missed about a third of in-bed breathing epochs, mostly in device-labeled wake.

Heart rate is the weaker radar claim

In that same aging paper, only the under-mattress devices (Withings, Emfit) produced heart rate. Somnofy did not. If a nightstand radar page shows a pulse number, ask which study backs that SKU. Do not borrow the mat numbers.

Stages are estimates

Sleepal's preprint is large (1,022 nights) and written by the vendor. Sleep/wake accuracy 92.8%. Four-stage 78.5% in the Normal (AHI at or under 5) group, 77.2% overall. It is a preprint, not an independent trial.

Chinoy and colleagues compared seven consumer wearables and nearables with PSG in SLEEP 2021. The through-line on this site and in that paper is the same: consumer devices are not a sleep study. If you already have a lab night, how to read a sleep study report walks the indexes. Wellness trackers do not replace that conversation.

Light vs deep disagreement shows up on other sensors too. Withings' own 117-adult home PSG comparison is the example we already use on the compared guide: sleep/wake can look fine while stages drift.

Komori Care (coming 2027)

Komori Care is not shipping, so there is no accuracy number to quote and we will not call it validated. It is a 2027 wellness product using 60GHz radar, with a back, side and stomach timeline as the design goal.

FAQ

Can radar measure heart rate? Less established for bedside radar. In PMC11387924 the radar measured breathing. The mats measured heart rate.

Is radar better than a wearable for sleep? No claim. Chinoy 2021 is the consumer-device context. Different sensors, different nights.

Why do stages look off? Light and deep are hard for consumer algorithms. A sleep/wake number can look fine while the four-stage chart does not.

What number should I trust? Peer-reviewed beats a vendor preprint. Say that about Sleepal's arXiv paper.

What about Komori Care? We do not publish accuracy numbers because it has not shipped.

Keep reading

• Under-mattress vs bedside radar
• Is the Sleepal AI Lamp worth it?
• Radar sleep sensing explained
• Sleep tracker accuracy claims
• Do sleep trackers agree?
• Best radar sleep monitor
• Sleep tracker validation scoreboard
• Sleepal vs Withings Sleep

---

Komori Care is a contactless sleep monitor in development. Camera-free position tracking, coming 2027. Wellness only. Join the waitlist.