How the Ultrahuman Ring tracks sleep stages

When we sleep, we cycle through three sleep stages — light, deep, and REM — alongside brief periods of wake. Each stage matters for rest, recovery, repair, and memory. Seeing how the night divides between them gives you a window into your recovery.

The Ultrahuman Ring reads three physiological signals that change as the brain moves between stages, and infers the most likely stage from the pattern they make together.

In this guide, we explain what those three signals are, how each stage is inferred, and how to use the results.

Key takeaways

• The Ring reads pulse, movement, and skin temperature, then infers sleep stages.
• The Ultrahuman app dashboard shows time spent in each phase plus deep and REM combined as your restorative sleep share.
• Read the result as a multi-week trend, not a nightly grade.

How the Ultrahuman Ring estimates sleep stages

• Heart rate: An optical sensor reads blood-volume changes at the finger with each heartbeat. From that pulse waveform, the Ring derives heart rate and the beat-to-beat timing that produces heart rate variability (HRV).
• Movement: A 3-axis accelerometer records how much you move and in which direction. Physical stillness is a precondition for any sleep stage, so this signal anchors the classification.
• Skin temperature: A temperature sensor tracks relative skin temperature at the finger. Skin temperature rises at sleep onset as core temperature falls, then follows an overnight curve that tracks your circadian phase.

A machine-learning model reads all three streams together and assigns every successive window of the night one of four labels: wake, light, deep, or REM.

How each stage is tracked

Each stage produces a different physiological pattern, which is read by the Ultrahuman Ring’s sensors:

• Deep sleep has the lowest heart rate of the night, minimal movement, and the strongest parasympathetic HRV signal [2].
• REM sleep keeps movement low, but heart rate timing turns irregular, and brief movements appear.
• Light sleep falls between the two and makes up the largest share of a normal night.
• Wake is inferred largely from movement. Lying still while awake resembles light sleep in the pulse alone, and motion is what separates them.

How sleep stages show up in the Ultrahuman app

The Sleep dashboard shows your restorative sleep share alongside your total sleep duration, with the individual phase percentages underneath. Restorative sleep is an Ultrahuman term for deep and REM combined.

Across more than 275,000 Ultrahuman Ring members (Ultrahuman internal data), the restorative share runs about 40 percent of the night in the 18 to 29 band and about 34 percent at 60 and over. Deep sleep moves from 17.0 to 14.2 percent across those bands, and REM from 23.3 to 19.5 percent.

How to use your sleep stage data

• Work in multi-week trends. A single night sits inside normal biological variation before measurement error is considered.
• Compare with yourself. Population averages describe a group. Your own baseline describes you.
• Lean on the restorative share as your main number, and read the individual phase percentages as the detail underneath it.
• Read it alongside how you feel. Where the data and the experience disagree, both are information.
• Use it to test changes, not to grade nights. An earlier last drink or a steadier wake time shows up over two to three weeks, which is the timescale the estimate supports.

Frequently asked questions

Does the Ultrahuman Ring measure brain activity?

No. It reads pulse, movement, and skin temperature, then infers the likely stage from how those signals change together. Measuring brain activity directly requires EEG sensors.

How does the Ring tell REM from deep sleep?

Both stages involve lying still, so the difference shows up in the pulse. Deep sleep has a low, stable heart rate with a strong parasympathetic HRV signal. REM has irregular beat-to-beat timing and occasional brief movement. It is one of the harder calls the model makes, which is why deep and REM are reported together as a combined share as well as individually.

Why did my sleep stages look different from how the night felt?

The estimate can be wrong, and it is least reliable on fragmented nights, after alcohol, during illness, on irregular schedules, around naps, and on very short sleeps — each of these blurs the signal patterns the model reads. Two other things also contribute. Brief awakenings you do not remember register as wake rather than sleep, and time spent falling asleep is often longer than it feels.

Do I need to tell the app when I go to bed?

No. Sleep onset and offset are detected automatically from the sensor signal [1].

Is a Ring sleep summary the same as a sleep study?

No. Polysomnography records brain activity directly in a supervised clinical setting and is a diagnostic tool. The Ring produces a nightly estimate for wellness purposes, read as a trend.

References

1. Driller M, Bodner ME, Fenuta A, Stevenson S, Suppiah H. How does the Sleep Regularity Questionnaire relate to wearable- and diary-derived sleep regularity? Journal of Circadian Rhythms 2026;24:2. DOI 10.5334/jcr.265 . This study assessed sleep–wake detection, not sleep stage accuracy. A total-sleep-time comparison against Somfit single-channel EEG appears in supplementary materials (4 participants, 20 paired nights). Devices supplied by Ultrahuman, which is named in the acknowledgements.
2. AlQatari AA, Alsunni AA, et al. Changes in heart rate variability and baroreflex sensitivity during daytime naps. Nature and Science of Sleep 2020;12:661–669. DOI 10.2147/NSS.S270191 (PMC7520661 ). 25 healthy men, PSG-recorded daytime naps, 9 of whom reached REM.