Face
Recognition
Facial recognition is the most common method (other than voice recognition) of personal identification (PI) used by humans. Because of this and the fact that facial recognition is non-invasive, usually passive, and fairly inexpensive, people generally do not have a problem accepting it as a biometric. It is probably for these reasons that face recognition has been one of the most active areas of biometric research. An example of the appeal and popularity of facial recognition is the mid-1990's Facial Recognition Technology (FERET) research program sponsored by the U.S. Department of Defense [1] [3]. This program was implemented to evaluate and advance state-of-the-art facial recognition systems and technology. As a result, there are now several companies that market commercial face recognition software that is capable of highly accurate recognition using sizable template databases (over 1000). The three prevailing technologies used in facial recognition are eigenfaces (or principle component analysis - PCA), local feature analysis (LFA), and neural networks. Depending on the application, a facial recognition system must be tailored toward three types of subjects: (a) a cooperative subject who is motivated to use the system properly because they might want to access an ATM, log-on to a computer, gain physical access through a portal, etc., (b) a non-cooperative subject who is either unaware or does not care that a system is in place and makes no effort to be recognized or avoid recognition, or (c) an uncooperative subject who actively avoids recognition and takes evasive measures [2].
The human face consists of a complex set of multidimensional visual images. As a biometric, it does not strictly adhere to the "permanent" characteristic. Non-permanent characteristics encountered by the facial recognition community include: facial expressions, aging, radical change of hairstyle or facial hair, eyeglasses, head and neck apparel, etc. Developing a basic computational model/method for facial recognition is a difficult task at best. When considering the non-permanent characteristics of the face (as a biometric) the complexity of the problem grows immensely. However, several technical successes over the past decade have made the possibility of PI using facial recognition appear both technically feasible and economically practical [3].
Facial recognition based PI is one of the most active areas of biometric research. Current facial recognition technology works well in "user verification" mode. For smaller databases, it works well in "user identification" mode. It possesses a biometric "permanence" characteristic deficiency that may require attention when considering this technology. Because facial recognition is (a) one of the primary methods in which humans identify one another, (b) non-intrusive, and (c) usually passive, users have typically accepted the technology with little dissension. There are three prevailing technologies available with each successfully implemented in several PI applications. A PI system developer should verify purported "user identification" capabilities before attempting a large user database application..
[1] A. Jain, R. Bolle, S. Pankanti, editors, "BIOMETRICS Personal Identification in Networked Society," Kluwer Academic Press, Boston, 1999.[2] International Biometric Group, http://www.biometricgroup.com. Last accessed: 30 July 2001.
[3] A. Pentland and T. Choudhury, "Personalizing Smart Environments: Face Recognition for Human Interactions," 8 October 1999, http://vismod.www.media.mit.edu/tech-reports/TR-516/ieee_computer.html. Last accessed: 30 July 2001.