Last Updated on March 16, 2022 by pg@petergamma.org
Newer Garmin watches
(e.g. Garmin Venu) can now measure respiration rate from wrist based optical sensors and newer Garmin HR chest straps (e.g. Garmin Fenix 6).
According to the Trainer Road forum, this seems also to work with the Sosche Rhythm+ heart rate sensors :
https://www.trainerroad.com/forum/t/respiratory-rate-from-garmin-hrm/44396/16
and eventually also with the Polar OH1. But are these data accurate and validated? If you look at the Matlab Respiratory Rate Estimation Toolbox, there are different graphs for ECG and PPG signals:
https://ch.mathworks.com/matlabcentral/fileexchange/55289-respiratory-rate-estimation
Are also different algorithms be used for ECG and PPG signals in Garmin watches? This would mean that signals from Sosche Rhythm+ or Polar OH1 might be faulty. This needs to be investigated.
Respiratory Rate Estimation project
One of the developer of the Respiratory Rate Estimation project is Peter Charlton, PhD, King’s College London. Algorithms for Respiraton rate for Garmin watches are from Firstbeat. Does Firstbeat use the same algorithms as Peter Charlton does? Or did Firstbeat reinvent the wheel? Firstbeat does not publish their algorithms, as far as I know.
There is a new project available from Peter H Charlton:
Peter H Charlton, one of the developers of the Respiratory Rate Estimation project:
http://peterhcharlton.github.io/RRest/
There is a new BITalino Mini Solo
available which eventually could be used for Respiratory Estimation:
https://bitalino.com/storage/uploads/media/bitalino-mini-solo-board-datasheet.pdf
BITalino Mini Solo is almost as handy as the
Firstbeat Body Guard:
https://assets.firstbeat.com/firstbeat/uploads/2016/02/Firstbeat_BG2_TechnischeSpezifikationen.pdf
Firstbeat uses as far as I know proprietary software, which cannot interface to Matlab or Python in real-time.
There is already a paper for the validation of Bitalino:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6498399/
They used Biopac Student Lab Pro software for the empirical validation of BITalino. Maybe a better choice for research projects is the Biosignalsplux Solo 16 bit data acquisition system: https://petergamma.org/biosignalsplux-solo-16-bit-data-aquisition-system/ Respiratory Rate from Garmin watches Respiratory rate from a Garmin watch during an activity? Is it accurate? How can this be validated? Counting the number of breath manually while running or cycling? Or Tracking Respiratory Rate with Pneumotrace in PowerLab during sporting activities? Even talking causes artifacts with Pneumotrace. This seems not to be an option. The Respitrace System
Respitrace is a research grade device. Respiratory Rate measurement is a research topic on which many papers have been published, but there is no gold standard device available, which can be used as standard reference device for all purposes:
Respiratory rate estimation for wearables project (Peter H. Charton)
Remarks:
1. Does it make any sense to remove motion artifacts from wrist based PPG watches for respiratory rate estimation? Why to choose a wrist based watch for this purpose? Which commercially available PPG watch allows access of PPG data?
2. Is it not better to use for instance a Polar H10 or Polar OH1 sensor with the Polar SDK, which allows access to PPG data and R-R peak data?
https://www.polar.com/en/developers/sdk
Garmin uses a newer chest strap to calculate respiration rate for instance with the Fenix 6. Garmin uses wrist based respiration rate calculation only at rest, for instance in the Garmin Venu:
https://support.garmin.com/en-NZ/marine/faq/2yEgS0Pax53UDqUH7q4WC6/
3. Polar OH1 and heart rate chest straps are scientifially validated and more accurate than wrist based PPG watches:
Why to choose a wrist based PPG watch for respiratory rate estimation? Is it not better to choose a Polar OH1 or a chest strap for this purpose, which is more accurate? These devices have an accuracy which is near to an ECG device.
23.1.22
The Garmin Venu watch
shows Respiratory Rate in real-time, or updates the display regularly, for instance in the Yoga protocol just shortly after you start the activity. I paired a Polar H10 to the Venu and did not wear the Venu on the wrist. The Yoga protocol showed heart rate, but no respiratory rate. Is wrist worn respiratory rate accurate? Peter H. Charlton PhD. says no because of motion artifacts during movements. But what is when the watch is at complete rest (altough not very comfortable to do for certain Yoga exercises)? Respiratory rate could be helpful for practicing breath meditation. Unfortunately, there is no gold standard reference device for respiratory rate for all kinds of purposes. I suggest to validate the device manually by counThis setup shows the complications that can occur with the PZT sensor as it is more prone to motion artifacts compared with the inductive RIP sensor.ting the number of breath taken in a minute and compare the data to the data on the device.
9.2.22
Which device to choose for calibration of respiratory signals?
Bitaliono respiratory belt sensors?
https://forum.bitalino.com/viewtopic.php?f=12&t=577&p=2249&hilit=respiratory#p2249
Peak detection for respiratory belt sensors seems to work
The happy sleep app shows respiratory beaks on the app, and I suppose they used beak detection or some other algorithms for respiratory rate calculation. At rest, the setup can be compared for instance to a Fenix 6 connected to a newer Garmin chest strap:
9.2.22
When I count the number of breath in a minute manually and compare the data to the data on the Fenix 6, the values on the Fenix 6 are around half of the values than the counted values. What is wrong? Is it the Polar H10? Garmin says, a HRM-Tri-, HRM-Run-, HRM-Pro- oder HRM-Dual- chest stap is required for Respiration rate. I currently only have a Polar H10 available. But do these chest straps not use standard protocols? Peter H. Charlton says, for respiratory rate calculation from ECG devices, the distance between two R peaks is required. Firstbeat says, they use heart rate variability information. But is this not the same? Is the R-R peak information or heart rate variablity information not also in the Polar H10?
9.2.22
No, it is not the Polar H10. I had the same deviation with a Venu wrist based respiratory rate, and a Taidacent respiratory rate belt with the HappySleep app. First I doubted about myself, but then I doubted about the devices. Fifty percent deviation from a manual count at a first test? For whom is this acceptable?
9.2.22
I had a serious crisis and went trough different youtube videos, checked the definitions, respiratory rate, breath rate, single breath, full breath, I could not find a problem there. I am really sceptical about the accuracy of Firstbeat products. When you for example compare the recovery time when you wear two different devices at the same time, there are big differences. The same with stress. Wearing two different devices at the same time, there are huge differences in stress values. The work of Peter H. Charlton is really of high quality, I think he did his PhD. about respiratory rate, and he was at one of the best universities in UK, and still is working for high quality institutions. There is a whole respiratory rate estimation project going on. From Firstbeat, I first want to see accuracy data before I trust these products.
10.2.22
An aternative to measure respiratory rate from Garmin watch is:
9.3.22
I ordered a Vernier Go Direct Respiration Belt from Vernier. The device is high quality, but very basic. The strap eventually needs a replacement by a more comfortable one.
A high quality device is also the Respiratory Rate Estimation Project by Peter Charlton. Unfortunately, there is no commercially available device yet there for this project, but a Matlab Toolbox.
A device which can be used for Respiratory Rate Estimation is Scott Hardens Sound Card ECG. I reviewed many affordable data acquisition systems. But I regularly come back to Scott Hardens Sound Card ECG as one of the best devices. It is alive, it is research grade, it is developed by an experienced biological research scientist, very affordable, and is supported by Scott on his Youtube channel.
Scott Harden Sound Card ECG:
Scott Hardens Sound Card ECG requires a bit soldering, but is very basic and very affordable. Sound Cards also have 16 bit resolution (research grade). Dr. Scott Harden, PhD. supports his device on his Youtube channel. Scott Hardens Sound Card ECG is also validated scientifically:
Validation of PC-based Sound Card with Biopac for Digitalization of ECG Recording in Short-term HRV Analysis
10.3.22
Now, I got the Vernier Go direct pro to work. You need a newer version of Android (Android 9 ? upwards). The advantage of the Vernier device is that it is ready to go. There are paper available which uses this device:
https://cs.nju.edu.cn/ww/papers/infocom21_wan.pdf
The Vernier does not need an additional respiratory rate toolbox, as for instance the Bitalino sensors do. Respiratory rate calculation is integrated in the Vernier Graphical Analysis software.
16.3.22
I had some issues with the Vernier Go Direct Pro Respiration Belt:
16.3.22
For a research grade respiratory rate measurement device with many publications, read the papers of Peter H. Charlton and his PhD thesis:
https://scholar.google.co.uk/citations?user=BJjD81oAAAAJ&hl=en
An accurate consumer grade Respiratory Rate Estimation Device would be highly desirable. Respiratory Rate Estimation is currently only available for Matlab, etc.
Matlab Home is a consumer grade application, but you are not allowed to publish with it. Who wants to buy Matlab Standard for 2000 USD a year for a respiratory sensor for research? Matlab is furthermore difficult for mobile applications. It does not run on Linux Phones or Linux Watches. Octave or Python are open source alternatives. I saw some Python code for Respiratory Rate.
Comments are closed.