As members of the U.S. Paralympic team begin competing this week in the Rio 2016 Games, Adam Tilton, CEO of Chicago-based motion-recognition startup Rithmio, will be watching the results closely.

That's because in the weeks leading up to the Paralympic Games, Tilton worked with Adam Bleakney, a former Paralympic silver medalist in the 800-meter wheelchair race and the current coach of several U.S Paralympic athletes, to develop an app to help athletes train more efficiently. Called Cadence Counter, the app measures an athlete's cadence, or number of arm strokes performed per minute in real time.

Tilton is hoping that, should the Cadence Counter app prove to be a fruitful training tool to Bleakney and his athletes, it will lead to more partnerships between the young startup and Paralympic and Olympic athletes. But the project also offers an inside look at the growing field of motion recognition and some of the challenges startups face in trying to adapt the technology to wearable devices.

Distinguishing between different types of motion.

Tilton, a former PhD student in mechanical engineering at the University of Illinois at Urbana-Champaign, launched Rithmio in 2014 with his co-founder and former professor, Prashant Mehta. As a student, Tilton developed algorithms that could estimate targets and classify patterns with precision for missile-guidance systems and satellites. They soon realized the pattern-recognition system they had built would also be perfect for the increasingly popular wearables trend.

Activity trackers like Fitbit and Jawbone use sensors to track metrics such as the number of steps you take in a given day. But it's a complex process to get the device to recognize that the motion you are making is a "step"--think about all the times when an app or a fitness tracker mistook a car ride as "walking."

To "teach" devices to correctly recognize the right movements, engineers typically use motion sensors to collect data from a large group of test users. Then they write an algorithm to recognize a motion that follows the general pattern of a step.

It sounds like a straightforward enough process--until you take into account how much variation there can be between people's unique movements, not to mention the fact that the process must be repeated for every single motion or metric an activity tracker wants to monitor. It's time-consuming, so it's no wonder that the majority of these wearables stick to monitoring simple activities like walking and running.

"We knew that we could [already] optimally extract useful information from sensor data, and then use it to make decisions," Tilton says. So he figured that they might be able to take those same algorithms and adapt them to human movement to track it more precisely.

They pitched Rithmio at the University of Illinois's new business-plan competition in the summer of 2014, and won. Soon, Tilton decided to take a leave of absence from his PhD work to instead dedicate himself to Rithmio.

Currently, Rithmio has 15 full-time employees. In June 2015, the company closed a $3 million seed round, led by Intel Capital. In the beginning of August, Rithmio released its first app, called Rithmio EDGE, in the Google Play store.

The app, which is specifically geared toward weightlifters, claims that it can "learn your form and automatically track your weightlifting workout." So you can show the app a bicep curl and it will start recognizing that motion and counting the number of reps you perform.

A Paralympic partnership.

Earlier this year, Tilton secured a meeting with Mounir Zok, the director of technology and innovation for the U.S. Olympic Committee, to try to sell him on Rithmio's technology and how it could help athletes. Zok encouraged Tilton to meet with Bleakney, also at the University of Illinois at Urbana-Champaign, to see if Rithmio could develop something that would work specifically for Bleakney's wheelchair racing and cycling athletes.

Wheelchair racers use the cadence metric to figure out not only what a certain speed feels like, but also to figure out the rate of strokes they might need to navigate different types of course conditions, like hills or a strong headwind.

"A coach might say to an athlete, 'I want you to hold a stroke count of 70 or cadence of 70 for 90 seconds, and then I want you to go to a cadence of 120 and hold that for 30 seconds,' and this is how they train," Tilton says.

Traditionally, Paralympians have just tried to count strokes or match movements with a metronome--both of which are tedious when you're just trying to focus on going fast.

"Cadence is a metric I've wanted all of my coaching career to provide insight on our athletes' form, not just the chairs," Bleakney said in a press release when Rithmio announced the partnership.

The Rithmio team started working with Bleakney back in July to develop the app. They took a week to interview the coach and his athletes about their training routines and the technology they previously used to measure results. They took another week to build the Cadence Counter app, using Rithmio's motion-recognition software called Constellation, and then spent another two weeks testing it with the athletes before giving Bleakney a finalized version of the app to use in Rio.

Industry challenges.

The motion-recognition industry as a whole is still pretty nascent. Dave Kim, an engineer at Toronto-based motion-recognition company Kiwi Wearables, explained in a blog post last year that there just hasn't been a lot of research done on motion recognition. He says that at the time of his post, the Institute of Electrical and Electronics Engineers (IEEE) had only 86,100 papers on motion-recognition technology for engineers to browse. That might sound like a lot, until you compare it with the 11.3 million research papers available on voice-recognition technology.

"Lack of research leads to less iterations of technologies for companies to use to improve their product," Kim writes.

Another hurdle for Rithmio is the fact that currently people wear motion sensors on just one limb, such as a wristwatch-style tracker. Tilton notes that that limb must be moving for the device to correctly register the motion. (This is especially tricky for leg exercises when you're wearing a tracker on your wrist--you're better off logging the exercises manually.) So Rithmio's initial reviews on Google Play are mixed because some users report that the app doesn't correctly count the number of repetitions, or incorrectly identifies an exercise.

The company also faces some formidable competitors, namely FocusMotion, a Los Angeles-based company that describes itself as the Siri for "movement recognition for wearables." The company has created a series of software-development kits (SDKs) that can track and understand metrics for a variety of exercises. This means that users can plug the SDK into any wearable and give it motion-recognition capabilities. The company recently received a seed investment from Kobe Bryant's venture capital firm, Bryant Stibel (the exact amount of the investment was undisclosed).

For now, Tilton is hoping the Cadence Counter will prove that the company can build software that can accurately distinguish between an even wider variety of motions. The Rithmio team is testing out the app on other athletes, including boxers, by placing sensors inside their boxing gloves and telling them to try to hold their fastest cadence for one minute, for example.

"The reason that this excites me is that you now have a very specific individual that has a very specific need from a wearable, and it's being met," Tilton says. "Now, we can step back and assess other problems this application solves for coaches, trainers, and athletes."