Thursday, February 3, 2011

What’s in a Face? Biometrics and Facial Identification

by Andrea Campbell 

We've all seen facial biometrics in the movies where a security camera picks out a terrorist subject or supposed perpetrator, reads their face and maps out a facial recognition identifier for law enforcement. It’s really flashy, cool and fast. Does it work that way? Well, yes and no. Let’s start at the beginning. 

Security and Biometrics

In the past, security measures have used a card, a token or key to get into closed doors. Other methods involve a password, code or something you must type in, like a series of numbers. The goal for biometrics is heightened security; meaning, it involves the identity of an actual person to gain entry. The word biometrics means it is some physical characteristic, or personal trait used to trigger an automatic recognition. We already use biometrics with fingerprints, a written signature, voice recognition or a retinal scan. Other lesser-used body differentials are someone’s gait, their ears, hand or finger geometry, or even their odor.


Several Keys

The key to biometrics' usefulness is that the characteristic must be measurable, and once presented to a sensor, can be converted into a quantifiable digital format. It is only good if it can be automated down to a number of seconds for retrieval. A system is called robust, if it can read traits that are subject to change. For example, the iris of the eye won’t change significantly over time, and it is more robust than someone’s voice. Plus, the more distinctive the identifier is, the better. A retinal scan would be more distinctive than hand or finger geometry.

Biometrics needs to perform two functions: identify and verify. In order to satisfy an identification it must work against records in a database, which it can search through for results, such as mugshots that are used by police. This comparison is called a "one to many" search and presents a best match result. Verification, on the other hand, is a system that relies on input from the user, generally via the password or another form of identity. This would be a "one to one" search, such as would be done with a computer allowing someone access using a private code. 

Did You Know? 


The police in Tampa Bay, Florida, used Indentix’s facial recognition software to screen at Super Bowl XXXV in January 2001. The search was for potential terrorists and criminals in attendance. The results? The system found 19 people with pending arrest warrants. Facial recognition is often able to operate with the used of surveillance cameras or closed circuit television (CCTV).
(Photo Courtesy of the Federal Bureau of Investigation) 

How Does It Work?

Facial recognition can be used to locate criminals, terrorists or missing children by using a five-step methodology. First, the image must be captured. This can be accomplished by scanning a photograph or using a camera to acquire a live picture of a subject. Fast moving video can also be turned into still shots. Next the software program seeks to detect the location of any faces in the acquired image, and it looks for two eyes and a mouth set into an oval shape. Once a face has been targeted, it is analyzed using spatial geometry. There are different methods available depending on the software, but this it is commonly referred to as the eigenface method. The system has templates that generate unique features for comparisons. That is to say, algorithms identify faces by extracting the landmarks of someone’s face and using the relative position, size and shape of eyes, nose, cheekbones and jaw. In other words, it measures the distance between the eyes, the depth of eye sockets, the shape of cheekbones, etc., but it actually ignores facial hair or hairstyles. These are then compared to a database of known faces, and finally, the scores are looked at by the end user for a determination. 

Better Than People


The software algorithms are generally better than people estimations. People are very good at recognizing people they know from surveillance film. With unfamiliar faces, however, not so good. The reasons for this are believed to be that people just aren’t good at it and they have a short attention span. To test this point, a British study used trained supermarket cashiers to screen shoppers. The shoppers were of four types. One had a shopping card with a recent photo, another had a card with modifications to their photo such as a minor hairstyle change or the addition of glasses. A third shopper’s card was issued that was actually of a different person who resembled them somewhat, and, finally, a shopper was issued a card where the only similar characteristic was the same sex and race as the shopper.

When the various cards were presented to the checkout clerks, more than half of the fraudulent cards were accepted. The breakdown was as follows. 34 percent of the cards that did not look like the shopper were accepted, 14 percent of the cards where the appearance had been altered were accepted, and 7 percent of the unchanged cards were rejected by the clerks. Time plays a factor too, as a human being's ability to detect critical signals drops rapidly after the start of a task, so that within 35 minutes their focus to task drops significantly. 

Current Uses 


A London borough uses a CCTV system, German Federal Police use it at a fully automated border station at an airport, Australia also has a system called SmartGate, and casinos and United States law enforcement agencies employ systems. It’s also been used at highly publicized trials, and to ensure fair counts for elections in Mexico.

The IAFIS (Integrated Automated Fingerprint Identification System) has an Interstate Photo File of 9,695,672 photos for over 5,629,772 records. The U.S. Department of State operates one of the largest face recognition systems in the world with more than 75 million photographs, and it is actively used for visa processing.

Results Are Weaker


The results of biometric facial recognition are still weaker and less efficient than the other methods, but police like it because it doesn’t require aid or consent from the test subject, and it can be installed in public places. 

Faces Are Public 


How do you feel about having your image captured, however? There is some controversy, but the privacy issue raised here is not protected by the U.S. Constitution. Under current law, the type of facial recognition used by law enforcement to monitor public places is legal because according to the United States Supreme Court, a person does not have a reasonable expectation of privacy because the physical characteristics used, such as one’s facial characteristics, voice, and handwriting, are constantly exposed to the public. 

New Technology 

A newly emerging trend, claimed to achieve previously unseen accuracies, is three-dimensional face recognition. This technique uses 3-D sensors to capture information about the shape of a face. You may also see technology developed as a security measure at ATMs, where a webcam image would be compared to a photo card.

When Madeleine McCann disappeared at Praia de Luz in Portugal, the British police asked visitors at the Ocean Club Resort to provide any photographs they may have taken in an attempt to identify the abductor of the missing child as part of the investigation, as some modern cameras have a focus and measure component.

Hey, even Facebook has a program that identifies faces in photos and allows you to tag other people you may know. I’m just saying, it’s going to be everywhere. 


Other Resource:
Bonsor, Kevin, and Ryan Johnson.  "How Facial Recognition Systems Work"  HowStuffWorks, January 2011.