My limited understanding of how Fitbit derive sleep state is through movement and variation in heart rate, they mention HRV changing as you transition through sleep stages.
This lets them calculate sleep stages with only an accelerometer and optical heart rate sensor and it appears their method of tracking sleep stages is between 75% and 80% as accurate as compared to a head mounted EEG.
So yeah, you don't need much to get a lot out. Would love to be able to integrate ML into the data processing, perhaps have a calibration period where you wear a sensor along with an EEG to improve accuracy through training.
Hey folks, I’m super new to Pine Time and wanted to figure out what I need to get started - would love to try and figure out sleep detection using accelerometer and HR data. What’s the simplest way to get it - should I buy a regular Pine Time or a DevKit?
Things I’d like to build:
- Sleep detection
- Adaptive alarm (wake me up if I experience REM between 7 and 8am)
Ideally I’d have the watch send the telemetry wirelessly to a server (maybe a Raspberry Pi) that will interpret the data and tell the watch if any action is required (eg buzz).
Does infiniTime offer this as is or do I need to hack it?
