Volume 33 Number 11
Editor's Note - Sensors in Sports: A Dive Into Applied Machine Learning
By Michael Desmond | November 2018
. It was a fascinating look at how ML can be used to improve outcomes in a physically dynamic environment.
This month, Ashley returns to our pages, exploring how sensor data and visual artificial intelligence (AI) are being used to improve the performance of athletes on the U.S. Olympic Diving Team. “Sensors in Sports: Analyzing Olympic Diving with Sensors and Vision AI,” details how Microsoft worked with coaches, athletes, and program leaders to deploy a sensor platform and smartphone app that lets coaches capture and analyze data related to athletes’ dives. At the time of this writing, the Microsoft team had just completed a live exercise at Stanford University with U.S. Olympic Diving coaches. The session showed how the combination of detailed telemetry and high-resolution video allows coaches to identify flaws in technique, refine training regimes, and improve outcomes in competition.
The effort got its start as part of a program at Microsoft to encourage organizations to integrate Internet of Things (IoT) sensors, data ingestion, ML and AI, says Ashley.
“We started Sensors in Sports as a collaborative effort at Microsoft to help our partners improve their integration of connected devices, technology and sports,” Ashley explains. “Understanding human movement is key to many industries. The models we build to analyze sports are useful to individual athletes, as well as our partners in many areas related to sports.”
Tracking sports movement comes with its challenges. Fast action means that remote sensors must operate at high frequencies, often at 100 samples per second, to produce useful data. Likewise, IoT algorithms must be autonomous and intelligent enough to keep pace. The physical sensors themselves must be hardened against environmental factors, such as water, ice and snow, as well as physical impacts. They must also be small, light and able to operate at low power, so their presence in no way disrupts training sessions.
“Over the last two years working with different sports, I’ve built about a hundred sensor prototypes, and now with the help of Microsoft Garage we can create even more custom designs for sports,” Ashley says, referring to the internal Microsoft program that supports what the company calls the “hack culture at Microsoft.”
Sports telemetry is a young and fast-moving discipline, and Ashley says that figuring out what data to capture, measure and analyze is a real challenge.
“This is precisely why we work with coaches and data scientists, training the best athletes in the world,” says Ashley. “Each sport is different, and at the same time, fundamentally, many activities are strikingly similar in how the human body reacts to the challenges. Our goal as sports data scientists is finding patterns, using ML to classify and analyze movements and find reusable methods in each activity.”
Michael Desmond is the Editor-in-Chief of MSDN Magazine.