Shame for Garmin accuracy for respiratory rate

Even hospital equipment has terrible respiratory rate accuracy. 

If you are interested in tracking your respiration rate, the Garmin watches are not doing great for sure.

For "sports" activities, the respiration rate is under estimated, but at least it is somewhat correlated to the actual respiration rate (r2 ~0.5 per the link you have), so the trends are OK.

For me the worst is that 3 activities, healthsnaphot, breathwork and yoga, are completely off.

During a single uninterrupted steady-state indoor cycling activity, I successively recorded 2mn sessions using various activities on the watch in random order to avoid a bias.

You can see below how snapshot (light blue), breathwork (yellow) and yoga (pink) are reporting a number way lower than "sports" activities.

Etupes25 said:

For me the worst is that 3 activities, healthsnaphot, breathwork and yoga, are completely off.

Great test!

It seems that sensor fusion (HR and accelerator) is used. This supports all activities with strong movement of the watch. Since healthsnaphot, breath work and yoga do not move the watch, the breathing rate is underestimated.

Didn’t the 10.33 update allowed chest HRM during breathwork?

forums.garmin.com/.../alpha-version-10-33

Yes. The test above is done with a chest strap. I also tested whether the presence of the chest strap was making any difference in the healthsnapshot. There was none.

It is as if the healthsnapshot, breathworks and yoga activities are constrained to certain respiration rates ranges.

- the healthsnapshot is always around the average breathing rate for an adult,

- the yoga activity is always slightly lower than the average,

- the breathwork activity is always the target, not the actual, respiration rate. Funny!

subra said:

This supports all activities with strong movement of the watch. Since healthsnaphot, breath work and yoga do not move the watch, the breathing rate is underestimated.

Not quite. Optical heart rate sensor do correct signal by filtering the noise coming from movement, yes. When there is no or little movement, there is less correction to do. When there is movement and the movement is a bit random (like resistance training), the correction is less effective: rhythmic movement (eg running, rowing, ...) is best at being corrected because it is easier to filter out the narrower spectrum of the movement signal.

Garmin Elevate Optical Heart Rate | Garmin

https://runningwritings.com/2021/05/cadence-lock-why-gps-watches-have-hard.html

Since the respiratory rate is estimated from heart rate variability, it unavoidably means that it's less accurate when the heart rate is low (= less heart beats for each inhalation and exhalation). Combine this with the difficulty of getting accurate heart beat timings from optical wrist sensors, and it's no wonder respiratory rate from low HR activities is not very reliable.

Remember that the "activities" above are just short recordings during the same biking activity.

So when you look at recording 3 to 5 for example, the indoor cycling (5) is the recording type matching the actual activity. For recording 3, i used a treadmill activity while I was biking indoor, and for recording 4, I used the health snapshot activity, again during the same indoor biking activity.

So, all activities should show the same respiration rate if they were really trying to show your actual respiration rate. It is obvious that something happens in the processing of snapshot, breathwork or yoga that prevents the real respiration to be reflected. Perhaps overzealous outlier correction based on "average" expected respiration rate during this type of activity. Hopefully the Garmin engineers will find out what is going on there.