[Pedometer] Pedometer is too much sensitive

[GiornoGiovanna2605]

Affected hardware version

Bangle 2

Your firmware version

The latest version available

The bug

###Describe the bug
It seems to me the pedometer is way too much sensitive and overstimates my steps number, even when I'm sit down it counts my steps

###Expected behavior
It should be more precise and count steps when I'm running or walking, not just when I literally move my arm

Installed apps

No response

[bobrippling]

Could be newer hardware, mine is very accurate - @gfwilliams wdyt?

[GiornoGiovanna2605]

I bought it not much time ago to be honest, like two weeks ago

Could be newer hardware, mine is very accurate - @gfwilliams wdyt?

[RKBoss6]

Hmm, for me, it overestimates my steps as well, but I've always found the difference to be too negligible to raise the issue... I got my Bangle this June.

[gfwilliams]

Around 4 years ago now there were some issues, but we got together a whole bunch of data from several different users (at https://github.com/gfwilliams/step-count) and HughB from the forum developed a step count algorithm that does a pretty decent job in most circumstances.

@RKBoss6 There was also some research done (https://pmc.ncbi.nlm.nih.gov/articles/PMC12074211/ and others) comparing the Bangle.js algorithm with various others, and if anything the Bangle appears to under-read slightly. It's basically an impossible task to get an algorithm that works perfectly for everyone.

@GiornoGiovanna2605 step counting works by detecting repetitive patterns in the watch's acceleration, so for instance if I were to sit, but rock my hand backwards and forwards repeatedly, it'd count steps - but so would pretty much any other step-counting device.

How many accidental steps are you seeing recorded when you're sitting? Is it a lot, or just a few? And have you been able to compare with another step counting device over the day?

Perhaps you could use the app at https://banglejs.com/apps/?id=accellog to record a few minutes of acceleration when you're sitting (and it's counting steps) and we could graph it here and see whether the accelerometer in your case was providing bad data, or whether there really was repeated movement that it was interpreting as steps?

[GiornoGiovanna2605]

I'll surely try the app. Normally it exceeds a lot in my opinion, last time I noticed this I had 123 more or less steps and after a talk with some friends while sat down I noticed it became like 180.

[GiornoGiovanna2605]

accellog.0.csv

I recorded my movements with the app and simulated normal gestures I would do in a discussion with someone. Steps went from 27 to 100+! Hope the file will be useful for you

[gfwilliams]

Ok, thanks - I just graphed it and got:

There are some big accelerations there! almost 2G in places (ignoring 1G gravity). By comparison I just recorded a very short period of me at a desk, with some minor typing and drinking a glass of water:

Not even 0.2G of acceleration.

Please could you record a period of where you're at a desk and basically not moving much? It could be something has happened that has made the accelerometer way more sensitive than it should be (but I doubt this as the base 1G of acceleration from gravity is still correct).

... or it could be that you just move your hands around a lot. But if that's the case then I think you'll find the vast majority of wrist-mounted watches you can buy will register steps too - they all work using acceleration of the watch as you move (either the swing of your arms or the bobbing up and down as you walk) and from their point of view, swinging your arms around while at a desk looks almost identical to how they swing around while walking.

To try and avoid logging steps when you're not walking, the step counter code looks for repetitive movements at about the rate of human walking (and ignores individual movements) - but looking at that acceleration plot the way the movement is being done, it its quite repetitive and so gets classed as walking.

[GiornoGiovanna2605]

So there isnt any solution for this? Is there a manual way to "pause" the pedometer to stop recording steps and resume it when I'm walking?

[GiornoGiovanna2605]

accellog.0 (1).csv
This is another example. When I was preparing my rucksack, it went from 0 steps to 25! Doesnt the watch have any methods to prevent this from happening? I wasnt literally doing a single step and it starter to counts the simple movement of putting my books in the rucksack as steps. Couldnt be developed some type of app that has an option to toggle on or off the pedometer? This would really help resolve the problem, because it's a bit of frustrating, to not know how many real steps I do

[thyttan]

I guess a workaround for that would be to start a recording of steps from the Recorder app whenever you want to count steps. It's even possible to resume a previous recording so you could just add on to it over the course of the day.

That will not affect the normal step counting of the watch but you will have the statistics you want through the recording at least.

It's also possible to modify the standard behaviour by tweaking the code - certsinly not impossible. You could have a special version implemented for yourself, or add it as an app.

If I may ask how come the precise steps are important? I'd think the step counting over a day or a week is probably not off by more than 10% or something like that (although not tested that myself). I always thought it seemed quite reliable - very precise while I'm walking.

[GiornoGiovanna2605]

It's just it drives me crazy that it counts more steps than the real ones, and I don't know a crap about code so I wouldn't be able to resolve the problem myself unfortunately ;(

[thyttan]

Ok! Here's the recorder at least if you want to give that a go: https://banglejs.com/apps/?id=recorder

edit: (There's an "Espruino work" section on the forum if you want someone to help you with the coding 🫡😉 Or if you want to learn there are a bunch of resources in documentation and tutorials on espruino.com. And the forum is there to ask questions if you want as well 😇)

[bobrippling]

I suppose one option would be to only count steps if (for instance) the step count clock info was visible, but currently the firmware unconditionally counts steps.

Then it'd be an effort to remember to turn it on/off each time you started walking. I think the real fix is to identify what's up with your accelerometer

[bobrippling]

As an aside, @gfwilliams is this a potential bug?

Should it be:

   if (a>32767) a=32767;
-  if (a<-32768) a=32768;
+  if (a<-32768) a=-32768;

I'll open a PR if so