Contributed editorial appearing in
Scientific Computing & Instrumentation 18:10, September 2001, pg. 16.
In his response to the 1997 Presidential State of the Union Address, Representative J. C. Watts stated “character is simply doing what’s right when nobody’s looking.” People of flawed character believe that a falling tree makes no sound if no one is around to hear it. In contrast, data acquisition declares the existence of an analytical signal before it reaches the detector. If such an analytical signal did not exist, signal modification devices including filters and lenses would not affect the detected image. Property privacy and security concerns must be considered whenever an analytical signal is broadcast in an open environment. There are laws that serve to discourage and punish illegal reception. However, the frequencies in the wave length range of 400 to 700 nanometers are generally considered public domain, as these are the frequencies detected by our sense of sight.We are constantly assailed by still and motion pictures appearing in newspapers, magazines, television, multimedia, and the cinema. Tabloid photographers earn a living acquiring snap shots of celebrities and public figures as they go about their daily lives. My picture even appears at the top of this page. The Federal Bureau of Investigation exhibits mug shots of the people on their ten most wanted list in post offices in the hope that citizens will report sightings to the police. Milk cartons and direct mailings display the faces of missing children and their suspected abductors so the general public can be on the lookout.
Visionics Corporation, headquartered in Minnesota and New Jersey, has automated the mug shot/eyewitness process in a system called Facelt. Visionics has developed a process that transforms a digitized picture of a human face into an irreproducible set of coefficients called a “faceprint.” The technology has grown out of automated digital fingerprint analysis wherein fingerprints discovered at a crime scene are digitized and compared against a fingerprint database for probable matches or a fingerprint scanner is employed to verify identity. The digital acquisition of a fingerprint, retina, voice, face or other feature used to identify an individual is known as biometrics. Visual biometrics uses digitized faces acquired using ambient lighting and a video camera.
The FaceIt software automatically searches a digital scene for the location of human faces represented by as little as 20 by 30 pixels. Both eyes must be visible and the pose must not exceed 35 degrees from the face normal. It then decomposes the face into a 12- to 40-coefficient faceprint through Local Feature Analysis (LFA). The LFA algorithm was developed by Visionics using a statistical analysis of characteristic features appearing in a representative ensemble of faces and is insensitive to lighting, eyeglasses, facial expressions, and hairstyles. The faceprint is
then compared to a Microsoft Access database containing known faceprints at a rate of up to 15 million faceprints per minute on a 500 MHz Pentium III computer running Microsoft Windows. In tests using the U.S. Army’s Face Recognition Test Database known as FERET, the error rate of false matches and missed matches was less than one percent. If the database contains the faceprints of wanted criminals or missing persons, the proper authorities are alerted to the match. If it instead is a database of properly authorized personnel, the system confirms identity and the person is permitted entrance, granted access to a bank account, allowed to vote, or logged onto the system, depending on the application.
The entertainment district of Tampa, FL, known as Ybor City, has recently distributed 36 cameras that are tied into the Facelt system. The police department monitors the matches found against a database of people with outstanding arrest warrants and can dispatch officers to the location. This application of Facelt has earned the scorn of critics who declare it as Orwell’s Big Brother come to fruition and fear the software will be used to spy on innocent citizens. In response, Visionics and the Security Industry Association have petitioned congress to create laws protecting against the misuse of the technology. At the forefront is a request for a “no match, no memory” policy requiring the destruction of captured faceprints if there is no match in the database. As our data acquisition and measurement prowess in technologies such as biometrics and the human genome increases, scientists, engineers, and business people must evaluate the moral, ethical, and public acceptance considerations when bringing their product to market.
