Vietnam's Viettel AI receives US patent for masked facial recognition technology

The U.S. Patent and Trademark Office granted the utility patent for the method, titled 'Method of Masked Face Recognition by Artificial Intelligence Technology,' Viettel AI said.

The company said its researchers developed the technology to improve facial recognition accuracy when users wear face masks.

Instead of relying on existing recognition methods, the research team built a large-scale dataset of partially obscured faces to train AI models to detect, search, and extract facial features more effectively.

The team also enhanced its facial feature extraction model and developed a high-speed similarity search algorithm to improve performance in real-world applications.

According to the company, the technology allows facial recognition systems to operate more effectively in applications such as attendance tracking, timekeeping, and access control without requiring users to remove their face masks.

It also supports the identification and tracking of people on watchlists or alert lists in public spaces even when much of the face is covered.

Viettel AI said the ability to recognize partially obscured faces was a key factor in meeting the U.S. Patent and Trademark Office's requirements for novelty, inventiveness, and practical applicability.

Facial recognition technology is widely used in applications ranging from attendance systems and device authentication to access control, but face masks have posed challenges for many systems.

According to a 2020 study by the U.S. National Institute of Standards and Technology, face masks reduced the accuracy of facial recognition systems, with error rates increasing by 20 percent to 50 percent on some platforms.