Bangle.JS - Poor HR Accuracy. Just me?

Hey!
My Bangle.js hr accuracy seems really low. I'm getting a (roughly) normally distributed curve hovering between ~90-150 bpm. But my other devices are reading much lower (~50-90). Most of the samples have low confidence and the high confidence samples seem to be multi-modal for some reason.
See the pictures attached.

Is anyone else seeing the same issues?
I've done multiple reboots and am taking the data sitting still at my computer. I get similar results when I'm not wearing the bangle. It's almost like the sensor isn't working at all.

Has anyone experienced the same thing or have any suggestions to boost accuracy?

2 Attachments

Nope, not just you :)
There are several threads for this. Best summary currently is in this post: http://forum.espruino.com/comments/16454­552/

Gordon has created a repository for collecting HR data to aid in the development of a better algorithm: http://forum.espruino.com/conversations/­373033/#comment16554946

Short version is, movement is bad for accuracy and currently for sports a Bluetooth based ECG-HRM device helps a lot.

@halemmerich: "ECG-HRM device helps a lot": good point.
Clear to us, but maybe not everybody is aware of bangle being able to receive, show and store the heartrate from external devices.
Added this to the summery.

I get similar results when I'm not wearing the bangle

Hi! Is this a Bangle.js 1, or Bangle.js 2? And which firmware version? I'd have expected a Bangle.js 2 to pretty much stop reporting BPM when it's not on your wrist.

When plotting the histogram, are you showing all BPM values, or the ones with a certain confidence? And also, are you moving a lot when recording the data?

With Bangle.js we take the approach that whenever there is a BPM reading we give it to the app, but with a confidence value. Honestly, if the confidence is 0, the value is likely complete garbage, and I wouldn't really be too convinced unless it's above 80. Even so your values don't look great though