For older people, falls are a major concern. It is the main cause of injuries to people over the age of 65, and the consequences can be life-threatening.
“That's not a problem in itself. It's a truly harmful injury to older people,” says Yejin Moon, an athletic science professor at Syracuse University who lost two of his grandparents to an injury that failed them.
The experience of losing family, friends, or neighbors from complications after falls is too universal. It's Moon and PhD student Reese Michaels G'24, combined cutting-edge research tactics (AI) video analysis with traditional laboratory research to learn how people will collapse and prevent serious injuries.
Analyzing analysis using AI and custom code
AI-driven tools such as OpenPose and WHAM have replaced traditional motion tracking markers, making it easier for researchers to study movement in real-world settings.
Traditionally, studying human movements has meant attaching motion tracking markers to the body. This is a common technique in games, film and exercise science. Today, however, advances in AI allow for direct analysis of movement from standard video footage.
“When you shoot video from your iPhone and enter it into the system, AI can automatically detect keybody points and track motion. No more markers are needed.
Working with Canadian researchers, the Moon and Michaels have access to over 1,700 real-life fall videos from surveillance footage from long-term care facilities and hospitals. Using OpenPose and Michaels custom code, the Research pair tracks body positions and extracts biomechanical data to assess which types of falls lead to damage and which movements protect against harm.
“It's like having access to a black box due to an accident,” says Moon. “You can analyze exactly what happened.”
Although Michaels had no previous coding experience, he took graduate-level Python courses through the Syracuse School of Information Studies. “It was a trial by fire, but I was able to write code for one of our projects, and I realized that I could apply those skills to meaningful research,” says Michaels, who began working with Moon at Folk Sports University two years ago as a master's student in exercise science.
“He can calculate autumn speed, acceleration, knee angle, or very specific biomechanical results, at the moment of impact.
As AI models continue to improve, team research will advance. “These new AI models can track movements in three dimensions rather than two dimensions,” explains Michaels. “It gives us much more insight into things like the angle of the joint during a fall, which opens the door to a more realistic and accurate analysis.”
“The goal is to implement this type of technology in a long-term care environment to gain real-time insight into how people move and injuries occur,” says Michaels.
Professor Yaejin Moon (left) of Fork College uses a special treadmill to simulate a sudden loss of balance, and a motion capture camera tracks participants' responses.
In the lab, AI models are validated using specialized treadmills that safely simulate balance losses. The treadmill can be moved forward, rearward, and left and right, but participants wear safety harnesses and adapt to sudden changes in movement. The motion capture camera records every step and response.
Falls occur in three phases: the initial stage (normal position or walking), the loss of balance phase (when the fall begins), and the impact phase (when the body hits the ground).
The new AI model allows researchers to track movement in 3D, greatly improving the accuracy and realism of fall analysis.
“Perturing treadmills are used to study the second phase of the balance at the moment they lose it,” says Moon. “We analyze how people react to losing their balance and how they try to recover.”
This study also investigates dual-task conditions. How does cognitive load affect your ability to restore balance? Participants are asked to perform mental tasks, such as listing animals while walking and counting from 100 to backwards per seven. This adds a layer of realism and simulates situations where thinking, speaking, and multitasking while moving can distract seniors.
“Does focus solely on walking quickly restore balance? The moon asks.
Research in the real world
PhD student Reese Michaels G'24 is the lead author of two studies. Scientific Report And it's currently under review Journal of Biomechanics.
So, how will this ongoing research affect people's daily lives? Moon divides it into three important components. “The first is to understand mechanisms. How do the body and mind work during autumn? The second is developing intervention programs. The third is improving technology.”
Currently in his second year with his PhD in Exercise Science, the Michaels program focuses specifically on improving technology.
Moon, the third-degree black belt in Taekwondo, began his research by teaching older people how to safely fall through martial arts. Now, she and Michaels are using AI tools to develop new ways to better understand falls and prevent serious injuries.
“One of our next steps is to feed a machine learning algorithm that can predict impact forces from a pose estimation model. “It gives us a direct measure of whether a fracture or damage has occurred.”
The pair also work to make video analysis methods more generalizable. With ongoing advances in AI and more real-world video data, the team hopes to analyze situations that cannot be replicated in the lab, such as going down a set of stairs, and address different ages and health groups.
By combining AI, biomechanics and real-world data, this study not only pursues fall research, but also lays the foundation for innovative solutions to prevent injuries in aging populations. As technology continues to evolve, their work will lead to more accurate strategies that can significantly reduce the risks faced by older people, and ultimately promise to improve quality of life and safety.
