How does facial recognition really work?
The Transformer 3 Pro can even tell twins apart—no matter how much they look alike.
With the laptop’s facial recognition technology, you can unlock your device in a flash just by smiling for the camera. The only password you need is that beautiful face your parents gave you.
So, how does this work? How does your device determine that its true owner is the one trying to access it and not some imposter?
The Transformer 3 Pro supports Windows Hello—software that covers a range of biometric logins, from fingerprint sensors to iris and facial recognition. There’s no need to type or swipe in a password manually.
Instead, the only information required becomes your biometric data—individually distinctive biological characteristics that can’t be faked.
Computer scientists have worked to develop facial recognition since the 1960s. At the time, this “man-machine” facial recognition relied on a combination of human effort and machine-aided precision.
Using RAND tablets, researchers manually marked photographs in places such as the corner of the eye or the center of the mouth. The tablets—early versions of the stylus and pad tablet system that is still familiar today—were notable at the time for their ability to sense the location of the stylus on the pad to an accuracy within one-hundredth of an inch.
From this data, scientists were able to calculate a set of distances for each photograph that could be matched against other sets of distances representing different faces.
Though we’ve come a long way since then, the core principle behind facial recognition remains the same: Breaking down the human face into coordinates and then matching them to a database.
When you set up Windows Hello, the camera on your Transformer 3 Pro captures your distinguishing features.
Today, scientists have determined that the human face can be split into about 80 different facial “landmarks” that outline exactly what you look like. These markers are known as nodal points and enable Windows Hello to map your face at set-up.
Measuring the relationships between these nodal points begins to create the unique set of physical characteristics—like the distance between your eyes, the depth of your eye sockets, the shape of your mouth, or the angle of your ears—that make your collection of features so indisputably your own.
Once captured, this data is converted into what’s called a “faceprint.”
You can think of it like a fingerprint: A physical identification that’s unique to each human. Your device converts your captured faceprint into ones and zeroes, before then encoding the data behind a dense layer of encryption—so that no one can steal it—storing your face as a blueprint for future reference.
This is a new kind of password made up not of letters and numbers, but of your eyes and ears and mouth and nose.
Each time you power up and look straight into the front camera, the software will compare what it sees to what it knows, searching your face for those 80 nodal points, and comparing them to what’s recorded. If you match the faceprint on file, the device is unlocked.
If, say, Mary-Kate Olsen would try to unlock her twin sister Ashley Olsen’s Transformer 3 Pro, she wouldn’t have much luck. She can’t just guess her password; she’d need Ashley’s actual face.
But what happens if Ashley presents a life-size, high-resolution picture of Mary-Kate to the camera instead?
Dual IR cameras prevent this dupe, sensing the depth of the object in front of it.
Relying on a high precision three-dimensional model of the registered user’s face (you will even need to tilt your head from side to side so it can get all the different angles and data points the model requires)—the software simply cannot be fooled.
Facial recognition is a quick, safety-enhancing feature—whether you’re logging in at home, or on a globetrotting, life swapping challenge half a world away.