[Help/Question] Garmin perhaps too sensetive

[seaside1]

I've tested multiple Garmin devices. Right now we have a Garmin Forerunner 255S. This watch is great for my kid, small in size, very good battery (I'm usually around 80% on the morning). Comparing this device with the Vivoactive 3, the 255S is a lot more robust (it easy even with ben-mode to accidentally exit the app on the vivoactive due to the touch screen).

The question is: Comparing Garmin with Pebble Classic, the Garmin is a lot more sensitive triggering a lot more false alarms, what setting is recommended to use for decreasing the sensitivity?

I'm also considering pros and cons with only looking at pulse and skip the shake alarm all together, or possibly buy a second Forerunner 255S but only measure pulse on that device.

[jones139]

Thanks for prompting me to think about the comparison between Garmin and Pebble. It is interesting that you can detect the difference.
The Pebble algorithm is basically the same as the one that we now use on Garmin, but included a few mathematical simplifications because of its low processing power, but we do the mathematically correct calculation on Garmin.

The main setting to reduce the sensitivity is the Alarm Ratio Threshold (default=57). Increasing it reduces the sensitivity. I have seen seizures with a value of 63 though, so do not increase it too much or the detection reliability will suffer.

If false alarms are occurring when there is very little movement, you can increase the Alarm Threshold setting to reduce those.

It is possible for me to analyse the effect of changes on the data in the Data Sharing database - if you have data in there I could do some analysis of your data for you if you'd like me to.

[pmithrandir]

Also, maybe check if the movements threshold is correct. I had to go to 800 to get something barely working by night. By day, it was not usable at all.

[seaside1]

Thank you for the response. I'll test this setting and see how it turns out. I usually run two watches in parallel so I can experiment on one. I don't have that much data to share, but I'll enable it moving forward. My child tends to shake a lot when waking up even without having a seizure.

Will also test the movement threshold, thanks pmithrandir (Using default of 5, but currently not using Neural Algorithm).

[jones139]

I definitely need to find some time to work on the machine learning algorithm - I think that is the key to improving false alarm rate - now dark nights are arriving I will try and get to it - I probably need to do another release of the Android App first to make it easy to download different models for users to try, then we can have a facility to choose which model you want to run...

[pmithrandir]

BTW, garmin comes with an auto lock feature. To unlock the screen, you need to hold one button around 3-4 sec.

Impossible to activate by mistake.

[seaside1]

BTW, garmin comes with an auto lock feature. To unlock the screen, you need to hold one button around 3-4 sec.

Nice, I have missed that, will take a look.