issues with the HRM

Ok, thanks! So with the app up to date that's looking a lot more reasonable, and it seems when confidence=100% the heart rate reading is matching the chest strap?

The picture above was what I get after a minute of not moving my arm.

And yes, movement will knock the HRM sensor off. There's a whole other thread on this which I think you'd probably have seen, but basically every watch sensor has this problem, and making it work better is a software issue.

• On modern sensors they look at the data from the accelerometer as well as HRM, and use this to try and discount false positives.
  
• The heart rate algorithm provided by the manufacturer for this sensor (which can use the accelerometer) is in a binary blob, with no source. Some people have been very against using this, so right now we have an open algorithm I came up with, but it doesn't use the accelerometer.
  
  

The code itself is at https://github.com/espruino/Espruino/blob/master/libs/misc/hrm_vc31.c (for the low lever sensor) and https://github.com/espruino/Espruino/blob/master/libs/misc/heartrate.c for the algorithm, so any improvements would be great.

BUT: We've been through this with the step counting. If someone makes changes that they believe improve things for them, it usually breaks things for other people. What we need is a more scientific approach - there's a thread here (sorry - don't have time to find it right now) where there's a discussion on this. Basically there's an app where you can download the raw HRM and accelerometer data and info from a bluetooth heart rate monitor - then we can get a bunch of data and everyone can test the algorithm offline and come up with improvements that work for everyone