Apple has received 73 new patents from the US Patent and Trademark Office. One of those patents details potential performance improvements for the company’s Face ID authentication platform that would allow Apple to support partial face recognition on its various devices.
As it stands, Face ID will reject an image and refuse to authenticate the user if only a partial face is detected during a facial recognition scan. The new patent would change that with a neural network that creates a bounding box when it determines that a face is present in an image. The feature would enable positive facial recognition based on a portion of the user’s face, regardless of its position or orientation within the frame.
The patent was originally filed in the third quarter of 2018, while its inventors are all members of Apple’s Machine Learning team. The new technology would also improve Face ID’s performance in adverse lighting conditions.
Apple has avoided partial face recognition in the past due to accuracy. The now-patented system would presumably give Apple a way to overcome that hurdle with Face ID.
If so, the new patent could theoretically allow Face ID to do a better job of recognizing people wearing masks. That, in turn, would make it safer to use during the COVID-19 pandemic. New York’s Metropolitan Transportation Authority has noted that many riders take off their masks to use Face ID, which increases the risk of spread in confined spaces like subway cars. The MTA has consequently asked Apple to update Face ID to reduce the threat of the disease.
In the meantime, Apple has picked up several other patents that involve Face ID in some capacity. One would introduce far-field recognition capabilities, while another would supplement facial recognition with vein pattern recognition during the scan. The latter would reduce the likelihood of false positives with Face ID.
Source: Patently Apple