US10719692B2 - Vein matching for difficult biometric authentication cases:
Subepidermal images of the user may be used to assess subepidermal features such as blood vessels (e.g., veins) when the device is attempting to authenticate the user. The subepidermal features may be compared to templates of subepidermal features for an authorized (e.g., enrolled) user of the device. Assessment of subepidermal features during the facial recognition authentication process may be useful in distinguishing between users that have closely related facial features (e.g., siblings or twins). In addition, assessment of subepidermal features may be used to prevent unlocking of the device by an unauthorized user wearing a mask or using another face replication method.
When I saw this show last month, the opening night pre-show curtain speech took a little longer than expected because the artistic director had all his notes for it on his iPhone... which he was unable to unlock with his mask on, something he openly complained about.
Reading the above, I'm guessing the phone was just waiting for him to get so frustrated that veins in his head popped up.
And here I was expecting a Nightbreed reference.
Close!
I imagine an enterprising lawyer is already drawing up a class action around a comparison to privacy rights around the photographing the outside/inside of your house.
... But only because I enjoy that kind of thought experiment at a very shallow level while I wait for the next shiny thing to pop up and distract me
This patent is from 2017. It describes FaceID.
No, it doesn't.
SFX make-up artists: creating masks with subepidermal features since [before this patent was lodged].
Are there any left? Fingerprint: tape. Palm: big tape. Iris: contact lens. Voice: recording/DF engine. Blood: those little stick-on latex pouches from that movie.
That doesn't work for this case -- those features need to show up to an infrared camera. Despite jwz's assertion above, facial veins show up in the IR image used for FaceID and are included in the features that are matched.
This patent is specifically about veins and blood, not Face ID in general, as is clear if you actually read it. The Face ID patent is 65230650 among others (which do not even mention veins and blood).
I did read it. I also worked at Apple at the time. Veins show up in the IR image used for feature matching.
That veins show up in the IR image has fuck all to do with the words that actually appear in this patent.
Are we reading the same patent? It describes the FaceID enrollment and device unlock processes. It describes using a VCSEL to emit speckle patterns onto the face for depth information, exactly as FaceID implements it.
In https://support.apple.com/en-us/HT208108, use of the IR image for feature matching is described.
All this matches my understanding of the technology from when it was being prototyped.
We are clearly talking past each other.
Me: "X is what this patent is about."
You: "Y is how Face ID works."
The two have only the most tenuous connection to each other.
How things actually work and what patents claim are very much not the same thing.
You do understand how patents are written, right? It's Maxwell Smart all the way down. "We claim this. As a modification to this. And if you don't buy that we also claim this. And would you believe an angry boyscout with a slingshot?" So yes, a patent about veins is also going to mention every single piece of prior art, in the hopes that something sticks to the wall well enough to not invalidate it. That first paragraph, "subepidermal features such as blood vessels" is what this patent is about.
Yes, I understand patents. I've got a few myself.
The claims in this patent continue to describe FaceID 1.0. This isn't the only FaceID patent, but it describes the IR-based feature matching portion fairly completely. There are other patents on other portions of FaceID such as enrollment and updating as well as elements like face detection, etc.
The primary inventor I don't know as he appears to be from a hardware team (HWE) but the secondary inventor, Feng Tang, is from the Vision team. They're responsible for ML-based work for various recognition needs. He's got several other patent filings for things used across the system like face detection, but also some for ARKit, such as RoomPlan, which was just announced this week.