Optical HR Accuracy – Making the perfect sensor for sports usage

Why isn’t there a Polar Optical HR Sensor? After all Polar have been the gold standard in sport HR monitors for a long time.

Source: Garmin, Forerunner 225

Source: Garmin, Forerunner 225

Why has it taken Garmin so long to pair up with MIO Global to include MIO’s optical HR technology in the new Garmin 225?

The answer is that these companies want to deliver products that are accurate for most or all people. The problem is that it is hard to do that.

Yet Garmin, Apple, Microsoft and many others HAVE taken the optical HR plunge – mostly as the market is demanding it and there is money to be made.

Edit: Polar announcement in August 2015.

We have heard about tattoos causing problems for the Apple Watch. Similarly darker skinned people can also experience problems as can people with a higher fat:muscle ratio. Sweat can cause problems and, indeed, how and where the device is worn can cause problems.

2x Epsons Optical Sensors

2x Epsons Optical Sensors

Optical HR is more accurate in the ear (JABRA and others) where there are less ‘motion artefacts’ caused by bones, arterioles and other ‘structures’.

IMG_3697

However the ear is not ALWAYS a practical location for a sensor for every sports person and movement of the sensor is possible. Further worsened by the fact that the ‘best’ wavelength of light to use is coloured green…and no-one likes green-glowing ears!

Scosche Rhythm+ HR Arm Band Strap

Scosche Rhythm+ HR Arm Band Strap

Another good place is the arm (SCOSCHE) but as we know from the Polar RCX3’s GPS armband from a few years ago; that’s not too practical a location either.

Source: Wellograph: 3x sensors (maybe, prototype)

Source: Swimmo: 3x sensors (maybe, prototype)

So we come to the wrist. There are LOTS of motion artefacts going on there; lots of movement to eliminate lots of veins and arteries and bones and fat (or not); as well as differing physical morphologies. BUT it’s a practical place AND we already wear devices there (watches!).

The location is obvious; the solution is not – otherwise it would have already been perfected.

A further problem with the wrist location is that if we wear our optical device TOO TIGHTLY then blood is pushed away; too loose, and there could be water ingress or light loss or too much physical sensor movement relative to the wrist location. This is not the same for chest straps where a tighter strap will still work.

adidas micoach - note the separation of the two sensors

adidas micoach – note the separation of the two lights

So the ‘perfect’ optical sensor (well a good one at least) may well need lot of fancy stuff that I don’t profess to understand BUT also two wavelengths of light will help to separate the good data from the ‘noise’ (see the Schosche image above 1xyellow+ 2xgreen light).

Samsung Gear Fit - Small Sensor Area

Samsung Gear Fit – Small Sensor Area

Another critical factor for the wrist location is the sampling area. If multiple readings are taken further apart then more chance exists to remove data artefacts.

The Apple Watch might have given this more consideration, as it is probably one of the areas where it’s accuracy is let down when engaged in sportier activities.

Matte Black - Back View

Source: Intel/Basis Peak – note the distance between green lights

Measuring Accuracy. MIO (Link, Fuse, Alpha, Garmin 225) state 99% EKG level accuracy. Yet if I wore two identical Polar H7 chest straps the accuracy/variation between them would probably be 97%!!. This is called ‘machine level accuracy’.

Wellograph - Image source: Gigaom

Wellograph – Image source: Gigaom

Whilst MIO’s claim, no doubt, is technically accurate it is a nuanced claim. 96% accuracy would actually be quite good (measured differently to the 99% EKG accuracy of course)  and, indeed, that is what other optical HR companies are aiming to achieve very soon and with that will be brought the tantalising possibility of HRV levels of accuracy-enablement for the wrist.

Anyway, we are expecting news today to explain why Apple’s WATCH has been having a few problems with optical HR accuracy at ‘sporty levels’ of exertion. Simply put: their sensors are probably too close together.

Edit: (HERE) is that news.

With that remedied and a better battery life/design the Apple WATCH 2 hopefully will be better.

Further exciting news in June 2015 sees Samsung’s SIMBAND (2x images below). The image is genuine and boasts a veritable plethora of sensors (many of the sensors are NOT for HR). The second image suggests wells-paced green sensor, much more wells-spaced than on any of the another released devices. Maybe this is the jump that is needed for the sports and medical accuracy we desire?

Source: ehealthnews

Source: ehealthnews

Source: smartwatchpro

Source: smartwatchpro

Optical Gallery

This slideshow requires JavaScript.

19 thoughts on “Optical HR Accuracy – Making the perfect sensor for sports usage

    • thanks for the comment Boris. I’ve never used the Scosche. dcr raves about it as we know. I’ve spoken to the MD of the chip manufacturer at valencell (who supply scosche) and they are pretty confident that they are nearly as accurate as chest straps. but then it comes down to usage and that’s where I agree with you. I used earbuds and they are perfectly fine for some other things and then once I start bouncing around the optical sensor must be moving relative to the skin and the signals are lost.
      also with the mio I’ve had pretty great results on one arm and then not so great on the other. I’ve got some of the more recent devices/firmwares from them at the moment which I have on my list of things to do.

        • If you had Polar H7 vs Polar H7 you will get about 97% accuracy !! This is machine level accuracy. Anyone claiming better than this must be ‘ill conceived’. so the graph that you showed is interesting as it is prob within similar levels of accuracy.

          • The absolute accuracy may be “good enough”. What I personally really disliked was the “jumpiness” of the optical HR reading. Also the tendency to always lower readings. was irritating.

          • the jumpiness will exist with any HRM depending on the granularity of recording – software will smooth it out. but I take your point. the drop at the start was interesting…that has happened with quite a few optical devices with me. strange it happens at the start. also at the end (sweaty, beating harder) there appears more error. really what would be useful would be to look at accuracy at various hr levels. if you read MIO’s university studies (not very good) then much of the figures are from quite low level sporting activity.

  1. may be based on the same stuff as here: http://www.valencell.com/white-papers I will read some time (maybe !). scientists can’t really argue with our real world experience. On the other hand though there are many issues with chest straps as we know. I know, for example, some women who simply can’t use chest straps and get no reading at all…anyway, point being we are all different. I’mw ith you for now tho that chest straps work for me, especially with the new garmin strap that, incidentally, I use with my wahoo pod :-). the newest garmin and suunto straps have 4x sensors unlike the older ones.but then the wahoo battery runs out too quickly…sigh !

    • I agree with you.

      (About battery: The battery of my H7 discharged very quick, also the battery of all my older Polar transmitters. At a daily usage scenario I had to change them every 3-5 weeks. Some month ago I switched to the TICKR, which I nearly always use with a simultaneous BLE and ANT+ connection. I still haven’t changed the battery after 2 month of nearly daily usage. Curious.)

      • hmmm. yes to be fair I get >2 months with Wahoo. Do you always at least disconnect at least one electrode? Polar good for AT LEAST 6 months . Maybe my heart beats real slow, giving the HR pod an easy life 🙂 actually it’s probably more because I use a variety of units. but garmin last for a year pretty much??

        • Yeah. Always at last one electrode disconnected. My weekly usage (means: amount of training time) hasn’t changed after switch to TICKR. I updated to the lastest firmware, as I got the unit. With the H7 I had sometimes drop outs at the beginning of a session. I’m using my old the Polar chest straps with the TICKR as I find them more comfortable. Had not one drop out since I have the device.

Leave a Reply