When your fitness tracker is not honest with you…
I know, it’s a long post, but it’s worth it! If you care about the accuracy of your fitness tracker, then you want to know about this!
Summary:
- Fitness trackers can be up to 30% off in distance and calorie estimation
- In both fields, mainly GPS causes this inaccuracy
- Barometric pressure sensors are able to compensate GPS measurement errors
- A fusion of accelerometer, barometric pressure sensor & GPS would be highly beneficial
Ever wondered, how accurate your fitness tracker really is? How good your smartwatch can estimate your calorie expenditure? And damn! Why is it telling me that I climbed 23 floors today? — I only took a walk around the city and my flat is located on ground floor level.
Well, first off: Yes, it is an estimation and companies do everything to make it as accurate as possible. They work on highly intelligent algorithms, trying to figure out when a step was made and how much calories were burnt. But there are limitations, especially when companies only implement a few sensors to estimate different values affecting your health.
A few weeks ago, I found an interesting article about the accuracy of fitness tracker, published by Which!. They tested several (85 to be exact) fitness wearable devices before the marathon in London. It was found that some fitness tracker are over- or underestimating the distance by up to 32%. This becomes important when you train for the marathon — I would guess, most people check the distance they ran on their fitness tracker — worst case, this would mean that during training, you run 8 miles more than required for the marathon. They also found, that one device’s calorie expenditure estimation was off by 30%.
How is this possible?
I recently published a blog post on the accuracy of GPS and how measurement errors can occur. Obstacles and clouds can block or distract the GPS signal. Loss of signal and therefore inaccurate distant measurement or calorie expenditure estimation are the consequences. For more on this, click here to go to the mentioned blog post.
Which! filtered the best and worst fitness tracker tested. You can view the list via this link. For this blog post it is not important to point out which wearable achieved the most accurate results. I want to look at another aspect — barometric pressure sensors…of course! After a short research on tear downs and the companies’ websites, I found that the three best wearables all have a pressure sensor included — the worst have not. Now, I am not saying that a built-in barometric pressure sensor can compensate an inaccuracy in distance and calorie estimation of over 30%. — but it definitely has an effect. Let me explain why:
Most algorithms base their calorie calculations on the data input shown in the figure below. I visualized the dependency of the considered parameters. This example looks at an average fitness tracker, with a GPS module, accelerometer and heart rate sensor.
The manual input done by the user is always included in the formula for calorie detection. For the formula to work, the user has to input and update the correct information. Highly important is the size of the user, due to the fact that the algorithm is deriving the step size from this input, which concludes in a measured distance and therefore the estimated calories burnt. If you have a watch with a GPS module, this could be beneficial for your calorie expenditure estimation. The watch gathers additional information, like the your pace or gained elevation. Walking up and down hills or staircases burns more calories than just walking on flat ground. However, as I already explained, the reliability of this information depends highly on signal strength. As I demonstrated in this blog post, sometimes altitude changes are recorded by GPS that never occurred, resulting in an overestimated calorie expenditure — and then again: What’s the point in monitoring your fitness, when the resulting data displayed to the user is inaccurate? Not much to say about the heart rate sensor: A built-in heart rate sensor is always beneficial for calorie expenditure estimation. Also here, of course it is important that it measures your heart rate accurately and best case, constantly.
Now, if you have a wearable with less sensors than this, you can see how the accuracy decreases. I myself, own a wearable that sports an accelerometer, GPS and GLONASS for tracking several parameters. However, using GPS results in a 98% drop in battery life time (In GPS mode, the watch’s battery lasts 10 hours, in smartwatch mode up to 3 weeks). So, you don’t want to activate GPS too often, except you are a big fan of charging your device every day.
And yes, you only activate GPS when doing sports, like swimming, running or playing golf. But the watch also tracks your calories, distance and steps when you are in smartwatch mode, which leaves us with a standalone accelerometer. Although, the device’s company is famous for its accuracy and well developed algorithm making sense of all the data recorded, in the end, it comes down to an accelerometer tracking your steps and deriving EVERY other parameter based on the steps detected.
Preliminary conclusion: As long as the user states the correct parameters regarding their gender, age, weight and size, the accelerometer and its related algorithm works just fine. But what can be done to improve the GPS-module’s accuracy? Right, adding a barometric pressure sensor — let me show you why this has a huge effect:
Elevation gain: A GPS module supported by a barometric pressure sensor can compensate GPS errors. For a better understanding, I want to look at the graph from my previous post again. In the figure below, you can see that the barometric pressure sensor detected a reasonable altitude. The GPS module shows changes of sometimes more than 10 meters that do not exist in the real environment the test was conducted. So imagine, calculating your burnt calories only based on GPS data — again, I am focusing on altitude here…of course, all the parameters in the above figure are normally also taken into consideration, plus: horizontal location determination via GPS is still the best available to the public currently — but including these wouldn’t change the fact, that the altitude tracking is way off, resulting in a wrong conclusion of calories & distance. Especially indoors, where GPS is not available, floor levels can be detected which makes it possible to show you the ACTUAL calories you burnt. With accelerometer- and GPS-only devices, this is mainly neglected.
Distance: To the left, you see one particular area in detail, where the pressure sensor and GPS differ by about seven meters in altitude. It’s safe to say, that the red line results in a much longer walk, then the blue line — walking seven meters up and down burns a lot more calories than walking straight for a few meters. And if you look at the overall graph, you’ll find a lot more of these examples — this was a test conducted over ONLY 10 minutes and there are multiple GPS errors — the only pressure sensor error is the influence by wind, which can be easily compensated, by double-checking if the accelerometer experienced an acceleration that justifies the sudden change in altitude.
Steps/Movement classification: Pande et al, is stating that in terms of energy expenditure, adding barometric pressure sensor data to accelerometer data increases the accuracy by up to 15%. Additionally a pressure sensor is able to enhance movement classification, as walking, stair/slope climbing or elevator use. Ohtaki et al showed in his publication, that the suggested caloriemeter, including an accelerometer and a barometer, provides results with higher accuracy than conventional accelerometer based calorie counter, as shown in the Figure to the left.
The Figure to the left stresses, that especially elevators cannot be detected by a standalone accelerometer device. However, adding barometric pressure sensor data to the algorithm, enhances the movement classification.
When merged, the lower graph shows the classification of activities experienced during the test. Therefore, an air pressure sensor would be a great addition to the calorie detection feature for fitness applications.
Oh, and last but not least — Why does your fitness tracker show you that you climbed 23 floors, when you actually know that you didn’t? Well, at least it means you have a barometric pressure sensor in your watch/phone that recorded the elevation you gained during the day. My take on this: Unfortunately, it also means that the company providing this feature, didn’t do the best job in communicating the value of this information. In my opinion, they want to make it more understandable for the user — they probably think that the user is not interested in an information like: ‘You gained 70 meters of elevation today.’ — or something like this. However, I think this is much more understandable for me. If the device tells the average user, that they walked more floors than they actually did, most people would say it’s not accurate. But what the device actually wants to tell you is: ‘You walked as much as 23 floors today — that’s 70 meters! Let’s do 25 tomorrow!‘. Makes more sense, right?
Thanks for reading!
Follow me for new stories on Medium.com or:
References
Journals
Ohtaki, Y., Susumago, M., Suzuki, A., Sagawa, K., Nagatomi, R., Inooka, H. (2005). Microsystem Technologies: Automatic classification of ambulatory movements and evaluation of energy consumptions utilizing accelerometers and a barometer, pp. 1034–1040, doi: 10.1007/s00542–005–0502-z
Pande, A., Zeng, Y., Das, A.K., Mohapatra, P., Sheridan, M., Seto, E., Henricson, E.K. & Han, J. J. (2013). Energy Expenditure Estimation with Smartphone Body Sensors. Proceeding BodyNets ’13 Proceedings of the 8th International Conference on Body Area Networks, 8–14. http://dx.doi.org/10.4108/icst.bodynets.2013.253699
Uniform Resource Locator
Which!, retrieved from: https://www.which.co.uk/news/2018/04/is-your-fitness-tracker-lying-to-you/
