Some random observations:
* My latest garmin has an "SpO2" sensor. Which is basically a random number generator with red LED. When I compare the results with an actual on-finger SpO2 monitor, there is zero correlation.
* Both the fitbit and the garmins have mistaken masturbation for deep sleep. You would think that the wrist movement would be a clue to not being asleep, but no. Deep sleep.
* The deep sleep vs REM classification seems to be based more on time-of-day than on anything else. Apparently I always start my nights with a block of deep sleep, followed by an alternating pattern of light & REM. This is true even when I'm awake during this "deep sleep" malarky.
* My CPAP, on the other hand, is very useful for telling apart sleep & waking. Breathing rate becomes much more steady when asleep. I can see when I woke up, even if it's just for a few seconds while I change position.
The developer stopped actively maintaining it, but the software still works great.
Looks like it's currently getting some TLC under the hood. Hopefully it'll continue to be maintained.
Perhaps someone more knowlegeable can pitch in?
More advanced ECGs can measure the impedance of the chest (which changes as you breath). This is more complicated to implement, but much more reliable.
In either case, it's not something you'd be able to measure without several stickers on your chest, with wires running to some central box. I think the CPAP would be less obtrusive. If you don't need a full fledged CPAP, you could use something like nasal capnography, that just measures exhaled CO2.
Though, it has recorded that I was in deep sleep while I was awake.