Enhanced Sleep Monitoring

[SunsetRunner]

I'm looking to develop an enhanced sleep monitor for Ionic that would be useful for learning about and tackling sleep issues, like breathing problems, bad dreams, etc. The basic idea is to wake up the user if irregular patterns are detected, so they become aware of what actually disturbs their sleep.

Since I'm totally new to the Ionic, maybe experienced developers or Fitbit team members can briefly let me know if the following is possible at least in theory:

1. Track heart rate (HR), blood oxigen (SpO2) and motion sensors all night

2. Wake up user at configurable events (e.g. HR high, SpO2 low, excessive motion, particular sleep stage entered, etc).

3. Use silent vibration alarm at wrist

4. Use an external Bluetooth speaker for alarm

5. Wake up user at best possible time in appropriate sleep stage

I'll be more than happy to start coding right away, if those are doable.

[JonFitbit · Accepted solution]

 There are some challenges to this:

1. Track heart rate (HR), blood oxigen (SpO2) and motion sensors all night

SpO2 is not currently available via an API

2. Wake up user at configurable events (e.g. HR high, SpO2 low, excessive motion, particular sleep stage entered, etc).

Sleep stages are only calculated after sync with site, and this data isn't available without querying the Web API. This would be too late for you.

3. Use silent vibration alarm at wrist

You can do this with the Haptics API.

4. Use an external Bluetooth speaker for alarm

Not currently possible.

5. Wake up user at best possible time in appropriate sleep stage

As above.

[SunsetRunner]

Thanks for clarifying, John.

Lacking EEG sensors and sufficient CPU resources for live processing, I already expected that deriving sleep stages from the Ionic's sensors alone would involve a lot of statistical matching and guesswork.

Provided there's enough RAM available, you might consider a technology that downloads with each sync some pre-cached data (probabilistic decision trees, or best: readily trained neural networks) that can be used to infer sleep stages offline with low CPU requirements. This would finally allow for the long sought after smart alarm feature, too.

Whether that is a viable path though, largely depends on AI skills available in your R&D team.

[paperinik]

@SunsetRunner, somebody is trying to fill the gap: http://bit.ly/2owOlfl

I can't comment further.

[paperinik]

My take here is that an approximation of the sleep stage might be already accurate enough for certain purposes; my experience with how this was implemented on the Pebble, which was not exactly a data center, was satisfactory. As for this happening consider that in general, this is one of the topmost feature requests and was marked "Under Consideration" on January 2014. Four years ago: http://bit.ly/2FE7OC7

I am simply stating the facts, I will let drawing the conclusions as exercise for the reader.