While you sleep, you go through a cycle of sleep phases. The first phase is light sleep, followed by deep sleep. A full sleep cycle lasts about 90 minutes and is normally repeated several times each night.
We deduct sleep phases from variation in decibels and sound classification in combination with some applied theories of the science of sleep. Here is a more detailed explanation:
We aggregate as much raw data as we can during the night. This includes all microphone input and device usage. The microphone input is interpreted as raw audio amplitude, and also classified as AI-recognized sounds.
When it comes time to calculate the sleep phase graph, we use a mix of sleep stage mapping to AI-classified recorded sounds (for example: Snoring means you are either in light, deep or rem sleep)
We also use ambient noise that peaks out from an ambient noise baseline (to detect non-classified noise, like getting up or moving from bed). This allows us to ignore constant white noise like air conditioners and fans.
Then, we use common sleep patterns & general sleep science data, mainly to try and estimate the non-awake sleep phases (it is not possible to accurately measure these with a microphone.) Even polysomnography, the gold standard, produces data that two sleep experts can disagree on.
Finally, we average all these weighted factors and produce a sleep graph.
Sleep score/quality factors in BetterSleep:
Sleep regularity (10%)
Sleep debt (45%) - hours slept compared to your goal or 7-8 hours average
Sleep interruptions (30%) - noise detected by the tracker
Deep sleep (4.5%), REM sleep (4.5%), and heavy snoring (4.5%)
Higher scores result from consistent bedtimes, sufficient sleep, and a peaceful environment.
Don't worry about low scores! Focus on feeling refreshed, practicing good sleep hygiene, and improving overall health.
We suggest this article from our blog: Poor Sleep Hygiene: Are you guilty of it?
Still have questions? Send us a message. We're here to help! 😊