LightBlog

mardi 26 janvier 2021

trinamiX’s new software paves the way for under-display facial recognition

trinamiX GmbH, a subsidiary of the German chemical company BASF, today introduced that its 3D imaging solution for secure facial recognition now works behind OLED displays. As demand increases for mobile devices with larger screens and smaller bezels, smartphone makers are turning to solutions that make the display transparent so they can move sensors like the front-facing camera underneath the display. Currently, the first generation of smartphones with under-display camera hardware does not support secure facial recognition, but trinamiX hopes its software will make this possible.

The company says its patented Beam Profile Analysis technology can detect if the person in front of the device is real, even if the person is attempting to bypass facial recognition by using a high-quality 3D printed full-face mask. The system runs data captured from a standard CMOS sensor and near-infrared light projector through the company’s material detection algorithm in order to detect real human skin. This material detection technology can also be trained for differentiating between other classes of materials as it is not limited to only skin classification. However, trinamiX is targeting its system on skin classification as there is a lot of demand for improving facial recognition, especially in our mask-wearing COVID-19 era.

While 3D facial recognition already exists on smartphones like the iPhone 12 and Pixel 4, it doesn’t yet exist on any smartphones with an under-display camera system. The ZTE Axon 20 5G, for instance, features an under-display camera, but the camera quality leaves a lot to be desired. “The challenge in making under display camera and face unlock work is that the cathode layer, a thin layer of metal that covers the entire surface of the display, absorbs a lot of light, particularly in the wavelength range required for the face unlock sensors to function,” explains Michael Helander of OTI Lumionics, a company which is developing a material it calls the “ConducTorr Cathode Patterning Material” for use in next-generation mobile devices with under-display cameras. A standard CMOS sensor and near-infrared light projector placed underneath an OLED display necessarily mean loss of light, but trinamiX says that its skin classification algorithm can still work with some loss in quality to detect real faces.

trinamiX says its 3D imaging solution is not ready for products yet, but the company is in talks with module manufacturers and smartphone OEMs to license its software. The company is also working with SoC vendor Qualcomm to ensure this technology works on Snapdragon platforms. In testing, unlock speeds were as fast as 0.2 seconds on a Snapdragon 855 test platform, with a false acceptance rate (FAR), false rejection rate (FRR), and spoof acceptance rate (SAR) of 1/1,000,000, 0.5%, and ≤0.1% respectively. The company says its technology supports Android 10 and above and runs on a Qualcomm secure environment — a necessity for storing sensitive facial recognition data. The goal is to bring this technology to the end consumer in 2022, we’re told.

Besides 3D imaging, trinamiX is exploring other sensor systems. Late last month, the company announced its mobile NIR spectroscopy solution. A small infrared sensing module collects data that is then processed by Qualcomm’s Sensing Hub in smartphones powered by Snapdragon chipsets, enabling on-the-go near-infrared spectroscopy. The company’s analytical models and “extensive know-how about molecules” are being applied towards scanning skin on a molecular level, at least initially, thus the initial applications of their mobile spectroscopy solution will focus on daily skincare. This technology has yet to be commercialized, though, and there’s also no word on when we’ll see the first smartphone with this solution.

The post trinamiX’s new software paves the way for under-display facial recognition appeared first on xda-developers.



from xda-developers https://ift.tt/39ZMyGs
via IFTTT

Aucun commentaire:

Enregistrer un commentaire