Exploring Ski Proficiency Estimation from Foot Pressure Patterns in a Simulator-Based Environment
This study demonstrates that a deep learning model using plantar pressure data can objectively classify skiing proficiency and track nonlinear skill acquisition trajectories in a simulator environment, while providing interpretable visual evidence linking forefoot pressure focus to improved performance.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are trying to teach a robot how to dance. You can't just look at its feet and guess if it's a pro or a beginner; you need to see the invisible forces it's using to stay balanced. This is the world of biomechanics, the science of how living things move and the forces they exert. In sports like skiing, staying upright isn't just about standing still; it's a constant, tiny dance of shifting your weight from your heels to your toes and from your left foot to your right. Scientists have long known that if you can measure these pressure changes, you can understand how well someone is moving. But what if you could turn those invisible pressure maps into a picture that a computer can "see" and learn from? This is where deep learning comes in. Think of deep learning as a super-smart student that looks at thousands of pictures, finds the hidden patterns that humans miss, and learns to tell the difference between a clumsy wobble and a graceful glide. The big question researchers have been asking is: Can a computer look at a map of foot pressure from a ski simulator and tell you if the person is a total newbie, a learner, or getting pretty good?
This paper tells the story of a team of researchers who decided to find out by turning a ski simulator into a giant, high-tech pressure pad. They didn't use fancy cameras to film the skiers' bodies or sensors on their helmets; instead, they focused entirely on the soles of their feet. They built a system that took the raw data of how hard someone pressed down on the simulator and turned it into colorful "heatmaps"—images where bright spots show heavy pressure and dark spots show light pressure. They fed these images into a powerful AI model called YOLO11, which is usually used for spotting objects in photos, but here, it was trained to spot "skiing skill."
The team gathered data from 27 people, ranging from those who had never touched skis to those who had skied for over 19 days. They asked the AI to sort these people into three groups: Beginner, Novice, and Intermediate. The results were a mix of "wow" and "not quite yet." When the AI was tested on the same people it had already seen, it was incredibly sharp, correctly identifying their skill level almost all the time. It was like a strict teacher who knows exactly how their own students move. However, when they tried to test the AI on a new person it had never met before, it got a bit confused. It was still pretty good at telling the difference between a total beginner and someone who had tried a few times, but it struggled to recognize the "Intermediate" skiers, mostly because there were very few of them in the group to learn from.
But the most exciting part of the story wasn't just about sorting people into boxes; it was about watching them grow. The researchers took six beginners and made them practice on the simulator over several weeks, measuring their feet every time. They watched to see if the AI's opinion of the skiers changed as they got better. And it did! As the weeks went by, the AI started classifying more of their movements as "Novice" or even "Intermediate" instead of "Beginner." To understand why the AI thought they were improving, the researchers used a special tool called Eigen-CAM. Imagine this tool as a highlighter that shows exactly which parts of the foot-pressure map the AI was looking at to make its decision. They found that as the skiers improved, the AI started paying more attention to the front part of the foot (the forefoot). This matched what the numbers showed: the skiers were naturally shifting their weight forward and becoming more active in their movements, just like real skiers do.
The researchers suggest that this method works as a way to track progress in a simulator, but they are careful to say it's not a magic crystal ball yet. The AI learned the specific "fingerprints" of the people it trained on, so it might not work perfectly on a stranger without more practice. Also, because the study only looked at people on a machine and not on real snow, the improvements they saw are specific to that simulator environment. Still, the study proves that deep learning can look at foot pressure, find the hidden patterns of skill, and even watch a learner's journey unfold, offering a new, objective way to see how we learn to balance on two slippery sticks.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.