Great, just great. I went relatively low-tech and outfitted the mattress I sleep on with an Angel Care sensor pad, connected to some custom homemade datalogging stuff. This is to convince me that what felt like a restless night actually did have significant sleep spells. So I'm just imagining feeling tired and having been awake most of the night, right? Have some coffee and get on with things. And now they're saying that sleep was, in fact, no good?
"Significant sleep spells" - sometimes the data, sampled in 1-minute binning intervals, is a flat line - no difference, minute-to-minute, in motion (however significantly above baseline of an unoccupied mattress - I guess that's what the angel care sensors monitor - just the motion energy of breathing. Other times the data is really, really noisy. Is that different sleep or just different body position? Never figured it out. But "awake" is clearly visible - that's noise that never goes down to the "just breathing" baseline at all.
If its based on heart rate and breath rate you could probably do that sonically or ultrasonically. If it's using ultrasound, it's probable that it's just because of your sleeping position. The maternal/infant monitors that the hospital my wife gave birth in used were very sensitive to positioning to the degree that they would stop measuring my son's heartbeat whenever she changed positions whereupon the nurse would come in and reposition the sensor.
Angel care sensor pad is just a piezoelectric disk and a couple of plastic plates. The real Angel Care box probably does DSP to pick out heartbeat or at least breaths, so it can raise the (frequently false) alarm if something goes wrong, presumably to trigger parent action that's fast enough to prevent a SIDS death. That's the theory anyway. In practice the false alarm rate is too high.
The algorithm I do use: Over a configurable number of samples, find maximum and minimum reading. Add the difference to an accumulator. Repeat. Once a minute, retrieve the accumulator and reset to zero. With 1-minute binning, it would be too slow to raise a "not breathing" alarm - but it's perfectly fine for post-analysis. I've used it for years, and can tell with a glance at the graph how bad the night was.
"Significant sleep spells" - sometimes the data, sampled in 1-minute binning intervals, is a flat line - no difference, minute-to-minute, in motion (however significantly above baseline of an unoccupied mattress - I guess that's what the angel care sensors monitor - just the motion energy of breathing. Other times the data is really, really noisy. Is that different sleep or just different body position? Never figured it out. But "awake" is clearly visible - that's noise that never goes down to the "just breathing" baseline at all.