Sports analytics—tracking how fast the ball is moving or how players move across the field—is becoming a key component of how coaches make decisions and fans view games. Data for these analytics is currently sourced through cameras in stadiums and courts and is incredibly expensive to acquire.
In an effort to make big data analytics more accessible for the sports industry, CSL researchers have utilized IoT devices—low-cost sensors and radio—that can be embedded into sports equipment (e.g., balls, rackets, and shoes), as well as in wearable devices.
“There’s a lot of interest in analyzing sports data though high-speed cameras, but a system can cost up to $1 million to implement and maintain. It’s only accessible to big clubs,” said Mahanth Gowda, a PhD candidate in computer science and lead author. “We want to cut down the expense significantly by replacing cameras with inexpensive internet-of-things devices (costing less than $100 in total) to make it possible for many other organizations to use the technology.”
The tiny sensors, which are wrapped in a protective case and distributed evenly in equipment, employ inferencing algorithms that can track movement to within a few centimeters. They can accurately characterize 3D ball motion, such as trajectory, orientation, and revolutions per second.
“This level of accuracy and accessibility could help players in local clubs read their own performance from their smartphones via Bluetooth, or school coaches could offer quantifiable feedback to their students,” said Roy Choudhury, an associate professor ofelectrical and computer engineering and computer science at Illinois.
The feedback could also help with detecting and analyzing player injuries, such as concussions. The sensor inside a soccer ball, for example, can measure how hard it hits a player’s head, giving coaches an indication about whether to treat the player for head injury.
The paper, to be published in USENIX NSDI 2017, explores tracking the 3D trajectory and spin parameters of a cricket ball; however, the core motion tracking techniques can be generalized to many different sports analytics.
The team, composed of students Ashutosh Dhekne, Sheng Shen, and other Intel collaborators, have also been developing methods to charge the sensors, including harvesting energy from the spin of the ball.
“We’re motivated to develop this technology to help coaches make better decisions on and off the field and provide enhanced entertainment to viewers,” said Roy Choudhury. “We want to bring advanced but affordable sports analytics to everyone, anywhere, anytime.”