Badminton, this fun and technical sport, which resonates both in school gymnasiums and on the Olympic stage, has nearly 200,000 members at the French Badminton Federation. But behind the apparent simplicity of this game hides a remarkable technique, difficult to acquire.
This is why, a team of Korean and American researchers revealed, in a study published on April 5, 2024 in the journal Scientific Data, a set of biomechanical data called MultiSenseBadminton, to analyze players’ practice. Their objective ? Develop training assistants based on artificial intelligence (AI) for driving enthusiasts.
“A shortage of datasets in badminton”
These personalized assistants, guided by AI, use cameras and sensors strategically positioned on athletes’ bodies to capture every joint movement pattern, muscle activation levels and even gaze movements. By analyzing this data, the AI offers personalized feedback on the player’s technique, as well as recommendations for progress.
And justly, “Badminton could benefit from these different sensors, but there is a shortage of comprehensive badminton action data sets for analysis and training feedback”explains Minwoo Seong, first author of the study, in a statement.
Read also: Corner, strategy, recruitment… What place for AI in football?
Rethinking badminton training with technology
Badminton requires precise execution of each shot. Badistas must demonstrate sharp reflexes, adopt perfect stance and master both the power and speed of their arm. So many elements that a coach can hardly correct simultaneously, in particular, for beginners who are not yet familiar with the basic movements of badminton, the acquisition of a swing posture and the use of appropriate power .
By using wearable sensors and AI to collect data from players of different skill levels, a system could be developed that would not only facilitate the training process, but also provide an objective measurement to complement performance assessment. a trainer.
Sensors for fine-grained movement analysis
The dataset proposed by the researchers captures the movements and responses of badminton players, helping AI-driven coaching assistants improve shot quality for all skill levels. Credits: SeungJun Kim at Gwangju Institute of Science and Technology (GIST)
In order to optimize these game parameters, the team of researchers from the Gwangju Institute of Science and Technology (GIST), South Korea, in collaboration with researchers from the Massachusetts Institute of Technology (MIT), CSAIL, in the United States, collected 23 hours of swing movement data from 25 players with varying levels of training experience.
Players had to repeatedly execute backhand clearing shots (a shot used to return the shuttlecock to the back of the opponent’s court using the racket on the opposite side of the player’s dominant hand), for example, or shots of “drive side rever” (a move which allows the opposing player to be moved laterally to open spaces on the field).
Sensors recorded their movements and responses such as inertial measurement unit (IMU) sensors to track joint movements, electromyography (EMG) sensors to monitor muscle signals, insoles for pressure exerted by the foot, and a camera to record both body movements and steering wheel positions. These sensors were not chosen at random since three coaches had brought their expertise to the team of scientists in choosing the most appropriate sensors for the design of a data set.
Read also:Motion capture reinvented for animals
“Personalized movement guides for every level of player”
In total, no less than 7763 data points were collected. Each move has been meticulously labeled based on shot type, player skill level, shuttle landing position, location of impact relative to the player, and sound at the time of the shot. impact. This data was then validated using a learning model to ensure its relevance in training AI models. MultiSenseBadminton dataset now available on online platform Figshareproviding a wealth of information to researchers around the world.
“The MultiSenseBadminton dataset can be used to build AI-based learning and training systems for racquet sport players. By analyzing disparities in motion and sensor data between different player levels and creating AI-generated action trajectories, the dataset can be applied to personalized movement guides for each player level. players“, says Minwoo.
In the long term, the researchers speculate that this dataset could make sports training more accessible and affordable for a wider audience.
2024-05-14 14:11:21
#progress #badminton
