Humans have a fantastic ability to deduce knowledge from visual observation. But human observation inevitably has limits. In the field of sports, judges and referees need to pass instantaneous judgment based on visual cues, which is prone to errors. In healthcare, therapists may advise physical exercises to their patients, but it is humanly impossible to monitor and ensure correct posture for all of them. Neither is it possible for patients to self-observe correctly. In retail stores with self-checkout systems, it is difficult for staff to manually detect and prevent fraudulent scanning motions intended to bypass the system.
In all of these scenarios, if human observation is complemented with AI acting as a trusted assistant, the ability to observe and make deductions increases several notches higher. AI-based, data-driven, judging systems ensure fairness and improve trust. Digital approaches allow patients to self-check postures. Gait detection can help detect early signs of dementia. AI-enabled self-checkout systems can automatically differentiate between normal scanning actions and fraudulent ones. When combined with the power of data and AI, this gives rise to a data-driven approach to analyze human motion at much higher levels of precision than is humanly possible.
Fujitsu Limited and the International Gymnastics Federation (FIG) marked a major milestone in their ongoing collaboration with the joint development and official introduction of the Judging Support System (JSS) for all 10 apparatuses[1] in competition at the 52nd Artistic Gymnastics World Championships in Antwerp, Belgium, held from September 30th to October 8th, 2023.
[1] Pommel horse, still rings, men's vault and women's vault, balance beam, horizontal bar, uneven bars, parallel bars, men's floor exercise and women's floor exercise.
The Fujitsu Human Motion Analytics solution is a data analysis platform that combines high-definition image analysis technology, know-how, and AI technology to analyze and visualize the skeletal motion of humans. The platform supports the systemization of human movement measurement and detection in multiple fields, based on high-precision skeletal recognition, a technology that has been field-proven for judging gymnastics performance. Data can be captured in up to 4 dimensions (3D + time series) to recognize motion as a series of movements. Besides being offered as an end-to-end solution with requisite licenses, installation, maintenance and support, the solution is open to partners to co-create smartphone apps and web applications focused on human movement.
Markerless motion capture
In contrast to common motion capture methods, the Fujitsu approach performs 2D or 3D motion analysis from videos captured by general-purpose cameras without the need to attach surface markers or motion sensors to the body. Besides motion analysis, interaction and relationship with objects are also detected.
High precision skeletal recognition
The Fujitsu Human Motion Analytics Platform is equipped with an AI learning model that can recognize a diverse range of agile and complex human movements with high precision, facilitating highly accurate analysis. The learning model incorporates an unique correction algorithm that significantly minimizes the jitter (estimation error) of posture recognition by deep learning, a common problem in image analysis.
Tools for rapid adoption
With the AI model pre-trained with about 100 basic human movements, and a large amount of artificially generated training data, the training period reduces significantly, enabling rapid adoption. Furthermore, the platform can be used even with commercial, off-the-shelf smartphone or web camera, significantly reducing the motion capture burden compared to conventional methods.
Human motion observations such as joint angles can be captured and analyzed as numerical data, and displayed in the same manner that judges need to see, based on the same standards.
The analytics platform can be used to check the accuracy of Pilates poses, comparing one’s own form with the ideal form, thereby achieving more effective training and improved customer satisfaction.
Human motion analytics is extremely useful in sports analytics. For example, in baseball, the service can be used to analyze players’ swings and pitching forms, correct wrong or difficult forms, analyze risk of injuries and prevent them, and improve overall player performance.
Gait detection is extremely useful for detecting signs of dementia and building an elderly-friendly society.
Human motion analytics is also useful to monitor and visualize the physical health status of workers and employees, ensure safety, and issue alerts or promote exercise habits.
With the increasing popularity of self-checkout retail stores, it is becoming imperative to analyze customers’ scanning motion to distinguish normal scanning with unintentional or intentional fraudulent scanning action.
Proven usage
Sports judging support:
The Fujitsu Markerless Motion Capture solution is based on the technology developed for, and in collaboration with, The International Gymnastics Federation (FIG). The system has a proven track record since 2019, with 95% accuracy and the results displayed within 5 seconds to less than 1 minute.
Healthcare:
Fujitsu MMC has been adopted as the gait analysis technology to detect signs of dementia by the National Cerebral and Cardiovascular Center, Japan.
Technological strength
Strong IP:
Fujitsu holds strong intellectual property including over 200 patent applications and has received considerable appreciation.
Awards and recognition:
Worldwide delivery
With more than 130,000 employees in 100+ countries, and partners worldwide, global-scale delivery and support is available.