The Face

Picture of a face.

Historical Information:

 The face was the focus point of the Bertillonage . Alphonse Bertillon created his namesake in an attempt to catch repeat offenders who would come into the police department under a different name each time. Bertillonage was capable of identifying people through beards and different haircuts by features that rarely change, such as the shape of the ear, etc. Bertillonage, however, had difficulties differentiating between identical twins and people who just happened to look similar. After time, Bertillonage was replaced by the Henry classification of fingerprints.

The Uses of the Face in Biometrics:

Face Geometry is not totally passive yet, since the person still needs to keep still in front of a camera. However, there are more advances in this area. Also, the face will sometimes change with time or injury, and that poses a problem. Some people believe that face geometry can be fooled by pictures. However, that is not true. Sometimes two camera are used at two different angles which allows the system to detect these pictures.
There is a wide range of costs for face geometry recogniztion technology. TrueFace created by Miros, for instance, costs only $59.95. It uses a camera that anyone can buy from a computer store to protect the computer from any intruders using face geometry.

Eigenfaces

Developed by Viisage Technology. First, a camera captures the image of a person. The face is mapped into a variety of coefficents depending on the features. Later the coefficients become 128 digits, otherwise called a "vector." The vector is the code for the person. When the person presents him/herself later to be identified, he/she must present a PIN or bar code and then allow the camera to take a picture. The difference between the face's vector and the database's vector for the PIN or bar code are placed on a coordinate system called "face space" and the differences between the two of them are determined. If distance between the images are small enough, access is granted. Otherwise, access is denied.

Local Feature Analysis

Local Feature Analysis (LFA) is another algorithim. First, the camera takes a picture of the face and cuts out the background and other faces. LFA is based on idea that all faces are made of building blocks. However, every face uses different building blocks and puts them together in a unique manner. LFA translates the building blocks into a code called "face print." The "face print" is stored in a database for identification. LFA is also used for verification. FaceIt by Visionics uses LFA.

Neural Networks

Developed by Miros, Inc.. Neural Networks is a slight bit more complicated than the "eigenfaces" method. A series of neurons, or processing units, are connected together. A programmer sets the rules of how the neurons recognize patterns. If the actual output is greatly different than the computed output, then the neurons will adjust itself to fit the situation. More situations lead to more shifts in neurons, which lead to better results. TrueFace by Miros uses this technology.

Back to the Head

Back to Home

Updates
Printer-Friendly Version