It sits quietly on your wrist: counting up your steps, tracking your sleep, monitoring your heart and calculating the difference between a light jog and a mad sprint. But how exactly does your fitness tracker come up with all the statistics that appear on the accompanying app?
Well, whether you’ve grabbed a new Boltt or maybe something like the Garmin Vivosmart 3, we got down and dirty with the tech and spoke to some of the companies trying to make sense of the data.
Starting off with the sensors
Simply speaking, fitness trackers measure motion: most of today’s wearables come with a 3-axis accelerometer to track movement in every direction, and some come with a gyroscope too to measure orientation and rotation.
The data collected is then converted into steps and activity and from their into calories and sleep quality, though there is some guesswork involved along the way.
Then there’s the altimeter that can measure your altitude, handy for working out the heigh of the mountains you’ve climbed or the number of flights of stairs you’ve managed to get up and down during the day. All of this information is collected and crunched to create an overall reading, and the more sensors your tracker has, the more accurate its data.
These sensors measure the acceleration, frequency, duration, intensity and patterns of your movement—taken together that’s a good bunch of data and it can help a tracker work out if you’re walking down the road or just waving at someone you know. Have a dive into the specs list of a particular tracker to see what sensors are included to collect data about you.
While the future of Jawbone is up in the air, its seasoned UP3 is still one of the most sensor-packed trackers, squeezing in temperature sensors and a bioimpedance sensor alongside the familiar accelerometer we’ve already mentioned. Bioimpedance sensors check the resistance of your skin to a tiny electric current, and the four electrodes on the inside of the UP3 fitness tracker are clearly visible.
Other wearables, such as the Fitbit Charge 2, use optical sensors to shine a light on your skin and measure your pulse through it: the light illuminates your capillaries, then a senor measures the rate at which your blood is being pumped (and thus your heart rate). These optical sensors are less effective than bioimpedance as a gauge of your overall health but can be more useful if you want to check your heart rate as you exercise or work out.
It’s a similar story with sleep tracking: using a process called actigraphy, your tracker translates wrist movements into sleep patterns as best it can, and as with steps there’s some guesswork and estimating involved. It’s a useful guide, but it’s not as accurate as polysomnography – this is what the experts use to measure sleep in a lab, and it monitors brain activity rather than how much you’re tossing and turning.
Add in the algorithms
As you might already know, it’s difficult to get two fitness trackers to agree on how much activity you’ve got through in a day or what your heart rate actually is. That’s because the sensors inside each device aren’t perfect at measuring what you’re doing – they all use slightly different algorithms to translate the raw data into actual statistics.
For example, your tracker might dismiss a small movement of the wrist and not include it as a step. But how small is too small? Different devices will have different thresholds and thus bring back different readings. Anything from a bumpy car ride to a plush carpet can throw off the accuracy of your fitness tracker.
When it comes to calories, an app needs more than just a step count to make the calculation: that’s why you’ll often be asked for your age, gender, height and weight too. The algorithms used by each manufacturer aren’t made publicly available, as each one likes to keep a lid on the ‘secret sauce’ they use to get the best and most accurate results, but the more sensors and data points used the more accurate the results are likely to be.
To really tell how many calories you’re burning, for example, a tracker needs to add data about your heart rate and how much you’re perspiring into the algorithm alongside how many steps you’re taking.
One of the most well-known tracking platforms out there is MotionX, which you can find in Nike’s running apps, the latest Swiss smartwatches and many other devices. Philippe Kahn is co-founder and CEO of MotionX developer FullPower, and he explained to us how the ‘signal processing’ procedure inside a fitness tracker cleans up the raw data that’s collected.
“Think about being at a concert. You just made a poor recording of a concert with a tape recorder of a great performance from a seat in the audience. Together with music, the recorder picks up all sorts of noises around you: your foot tapping, chatter, interactions… if you wanted to turn this poor recording into a quality recording of the music, you’d have to eliminate as much of these unwanted sounds as possible.”
And the quality of this cleaning up process varies from tracker to tracker, Kahn says. The applied methodologies range from the simple to the sophisticated, but MotionX goes all in: Fullpower uses more than 100 specialist engineers to tweak the accuracy and effectiveness of its software, and the company has invested more than $50 million in the development of next-generation algorithms — this is big business.