The Complete Beginner's Guide to Face Recognition....

The Complete Beginner's Guide to Face Recognition

Face recognition working

Face recognition is the process of identifying or verifying the identity of a person using their face. The smartphone industry is adding different methods to unlock your phone out of which face recognition is relatively new. Face unlocking debuted in Galaxy Nexus, which came along with Android 4.0 Ice Cream Sandwich in 2012 it just used a front-facing camera due to which it was not very secure and neither was it marketed as a major flagship feature.

But as he has gone by, face recognition got so much better and companies like Apple started marketing it as a flagship feature claiming that it will be more efficient than the fingerprint scanner or Touch ID. The face ID in iPhone X, face unlocking in Oneplus 5T, and facial recognition with iris scanner in Galaxy phones.

So, let's know-how exactly face recognition systems work.

Most systems work by creating patterns on the face and then identifying them just mimicking the fusiform face area in the human brain. The system divides the face into visible landmarks called nodal points like eyes nose and mouth. It captures the light emitting from its own light-emitting diode which touches the face and reflects back are the natural light which reflects from your face and then identifies this nodal point by colors.

For example, it can identify eyes with a combination of black and white colors and if any nodal point is missing the system won't work. This is the reason why companies claim that face ID won't work if you're sleeping or not looking at it.

Further, the algorithm uses complex techniques to identify whether the light is receiving from the nodal points are natural light or the light emitted from its own source so that it cannot be tricked by showing other sources which display your face the algorithm also calculates the time taken for light to travel from each nodal point to the sensor.

For example, if your eyes and nose are two different nodal points the time taken for light to reach from your nose would be less than that of the time taken for light to reach from your eyes This way, the system would be able to identify whether it is looking at a three-dimensional surface and hence cannot be fooled with the photographs.

Once a person calibrates his face to the facial recognition system the system gets information from different nodal points and combines them to create a unique face code or face print. While validating any face it tries to match the face print of that face with the calibrated one to unlock your phone with some margin of error.

Now using these systems have five major problems and solution is also there:

1. The aging problem happens as there are changes in features of the face due to which the algorithm may not be able to recognize the same calibrated face in just a matter of a few days.

To avoid this, algorithms nowadays use rigid tissues of the face as nodal points which won't change as time goes by.

2. The same applies to emotions also as algorithms use those facial features which won't change during any emotions as nodal points.

This somewhat ensures that the algorithm recognizes your face whether you are crying at it or smiling at it.

3. The problem of illumination happens for those systems which use only the front-facing camera like the Oneplus 5T as they depend on external light sources and cannot be used in darkness.

This problem is solved by some systems like the one in iPhone X by using their own light sources and sensing the same once it gets reflected back from the nodal points.

4. Regarding the position problem if the system makes a 2-dimensional faceprint of the calibrated face it may be possible that the face may never come at the same position and angle as the calibrated one.

To solve this while calibrating any face the system creates a 3d model of the face being calibrated so that it can evaluate the same at any angle and at any position.

5. Last but not least the twin problem is basically, the problem that any identical twin can trick the system and unlock the phone.

Samsung's iris scanner solves this problem by using a special sensor to identify the patterns existing in human eyes. This is more secure than the fingerprint scanner as each human iris has a distinct pattern and even the left eye and the right eye of the same person have different patterns. The twin problem is pretty much unsolved in other facial recognition systems.