Live demo
One real recording, scored by a human technician, run through the engine. Every number on this page is read from the engine’s output file. Nothing here is illustrative.
Total sleep time
5h 27m
ngt_sc4001e0 · 7.0h recorded · 30s epochs
The hypnogram is what a human technician wrote down: one of five labels per epoch. Underneath it, on the same axis and the same minute, is what the EEG was actually doing. Structure that the five-state scoring flattens stays visible in the continuous series: a shallowing N2, or an arousal that never reached wake.
One shared time axis, ticked in h:mm elapsed from lights out. Above: the technician’s five-state scoring, one step per 30-second epoch. Below: Odds Ratio Product, a continuous 0–2.5 depth estimate, where low is deep and high is awake. The colour strip is the same ORP series binned to the plot’s own pixel resolution (mean of real samples per bin); bins with no valid sample are left empty rather than filled. Unscored epochs and null samples break the line instead of being interpolated.
841 scored epochs · 841 valid ORP samples of 841
How the night divides across the five scored states.
Percentages are of 7h 01m of scored recording. Depth stages share one hue stepped by depth; REM carries the only second colour because it does not sit on that continuum.
Each card names the paper it implements. Where a metric cannot be computed from this input, the card says so and says why. Nothing is rounded down to zero and nothing is estimated. Each card also states what its value describes. Most are computed from this night alone. The Sleep Regularity Index is subject-scoped: it is computed across subject 0’s consecutive recording days, so it is identical on every night that subject contributed.
This night
Reference classW+N1, widened from the published wake-only class: on this 8-night corpus a wake-only reference leaves the sleep end of the scale unpopulated, collapsing N2, N3 and REM to indistinguishable values
Continuous sleep depth on a 0–2.5 scale, derived from deciles of 3-second EEG band power. Low is deep, high is awake. It resolves gradations inside a single scored stage.
This night
Sleep spindles detected per minute of N2 sleep.
Lacourse et al., J Neurosci Methods 2019;316:3-11 (A7)
doi:10.1016/j.jneumeth.2018.08.014This night
The overnight dissipation rate of delta power across successive NREM periods. In the two-process model this is the discharge of Process S.
This night
The power-law exponent of wake-bout durations across the night (reference value ~1.3), reported alongside stage-transition entropy.
Subject-scoped · subject 0
Measured across 2 consecutive recording days for subject 0 (46.6 h wall clock), not from this night alone. The same value therefore appears on every night this subject contributed.
Percent concordance of sleep/wake state at time points 24 hours apart, rescaled to 0–100. It is measured across multiple consecutive days of recording rather than within a single night.
Phillips et al., Sci Rep 2017;7:3216; Windred et al., SLEEP 2024;47(1):zsad253
doi:10.1038/s41598-017-03171-4Provenance
These are real polysomnography recordings from Sleep-EDF Expanded, published by PhysioNet. The hypnogram is the original scoring by a human sleep technician, not a Beacon prediction. Beacon computes the signals plotted against it. No values on this page are simulated, smoothed for presentation, or filled in.