Running Motion Capture with Xbox Kinect V2 – Why is it interesting?

According to the American Academy of Physical Medicine and Rehabilitation, nearly 70 percent of runners will become injured. Various studies peg the percentage of runners that will get hurt in a given year at 20-80%, which is staggering. Runners are told to pick shoes currently, get orthotics, not ramp mileage too fast, take time off, to cross train, lift weights, and so one.

To the familiar eye, however, the seeds of many overuse injuries can be seen in a runner’s mechanics. The unstable left ankle seen on a run precedes a comment about a tight left Achilles. The knee valgus is observable well before IT Band syndrome sets in. But for the average runner whose most sophisticated gait analysis comes from a shoe store salesman, these things are unknown. But even if the problems were known and a regimen of exercises was prescribed to fix them, how would the runner know he or she is improving? After all, what gets measured gets attention and what can’t be is often ignored.

What if there was an affordable tool that had the ability to record how one runs and gives feedback over time to continuously eliminate weaknesses? What if a video game console in millions of homes nearly has the technical capability to do this?

Above is running motion captured on an Xbox Kinect V2 sensor using its 3D “image capture” and skeletal tracking capabilities. The recording was done with the help of the second generation Microsoft Kinect Studio software, right off the shelf. I am fascinate by the possibility of using technology that is this inexpensive ($149) to capture human movement and provide feedback to athletes on how they are moving. It gets incredibly interesting if the solution can reach athletes AT SCALE. I see the key to a compelling product is one that (1) provides accurate analysis of motion (2) in the setting where the athlete trains (3) at a reasonable cost (4) over time (5) in a way that provides actionable insight.

The Kinect solution is nearly there in two of these aspects, and the other two will take some work.

  1. The accuracy of the Kinect V2 sensor tracking is good, better than the first generation, but not great, especially with fast motion. Short of having a high-precision truth system to compare the joint positions, the main evidence of this is the joint noise, apparent when a joint rapidly changes to an unrealistic position when the sensor loses sight of it.
  2. The Kinect is an excellent form factor that can easily be moved and setup where the athlete trains.
  3. Only requiring the $149 sensor and a PC, it’s a system that can can be set up most places athletes train. It’s important to note that the Kinect’s reliance on IR light dot clouds does prevent it from being used outdoors, and makes using multiple sensors at the same time difficult.
  4. The requirement of recording several data sets over time is the most compelling feature of a system like this. Data from motion analysis is useful in a single assessment to identify the athlete’s areas of strengths and weaknesses, but things get really interesting when motion or metrics can be compared across days, weeks, months, or years to really know if the athlete’s training regime is improving their performance and mitigating injuries before they happen. The mental, social, and financial benefit of preventing even 10% of injuries in athletes would be huge. High-end systems in research universities and specialized clinics are great, but your average athlete won’t be able to afford a motion analysis on a regular basis.
  5. And the last ingredient is creating a system that provides actionable insights about the motion captured, whether that be metrics generated, an ecosystem of virtual motion analysis professionals, etc. This is a really tough software and services problem, but it is definitely solvable.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s