NEC has upgraded its facial recognition engine to make sure that it can identify people who are wearing masks. The new engine was able to match faces with more than 99.9 percent accuracy in internal 1:1 testing, regardless of the color of the mask and any patterns printed on it.
The engine focuses on the exposed areas around the subject’s eyes in order to achieve those results. It is also able to determine whether or not an individual is wearing a mask at all, and will automatically switch to the appropriate algorithm based on its conclusion. The original algorithm looked at the mouth, the nose, and other features in addition to the eyes.
The update is a response to COVID-19, which has encouraged the use of masks as people search for ways to slow the spread of the disease. With the new engine, people will not need to remove their masks to use the face-based identification systems in mobile devices, work stations, and other venues, which could lead to better public health outcomes overall.
Since it leverages multiple algorithms, the new engine can be deployed to scan a mixed group of masked and unmasked individuals. NEC will begin selling to customers later in the fall, when it will be available in products that offer “NEC Enhanced Video Analytics” and “Bio-IDiom Services.” The former is a more general computer vision and video analytics program, while the latter pairs face recognition with other forms of biometric identification. However, the Bio-Idiom Services will only be available in Japan for the time being.
NEC’s facial recognition algorithm has been listed as the top-performing solution in multiple rounds of NIST testing. The company has already released a separate mask identification solution that combines elements of face and iris recognition, and has argued that COVID-19 will generate more interest in contactless authentication technologies.
September 25, 2020 – by Eric Weiss