Ўзбекистон республикаси ахборот технологиялари ва коммуникацияларини ривожлантириш вазирлиги муҳаммад ал-хоразмий номидаги

 
A BIOMETRIC ACCESS CONTROL SYSTEM BASED ON THE HUMAN 
FACE 
M.R. Raxmanqulova (master,TUIT named after Mukahmmad al-Khwarazmi) 
Fraud is faced on a daily basis in today's society. An impostor could slip 
through and access highly sensitive locations, a child could be released to a 
stranger from a daycare center, a welfare recipient could sign up for benefits under 
six identities, a voter may vote under two different names using two different 

---

350 
social security numbers, and a counterfeiter could make copies of and charges to 
credit cards. Authentic means of identity verification are urgently in high demand.
Biometrics, which is the field that incorporates the measurement of one or 
more distinctive biological trait(s) in order to be studied, examined, or used to 
uniquely identify its owner, is the most promising solution. A biometric system 
that employs the human face for identity verification and access control was 
successfully developed and tested. In the course of achieving this goal, two 
different models were developed and tested. One employs a neural network and the 
other uses feature extraction by implementing a novel elastic template matching 
approach. 
Biometric-based identification systems, which are defined above, can be 
characterized either according to the biometric traits used or by the type of 
identification method. Characterization based on biometric traits is considered first 
and characterization based on the identification method is considered next. 
Figure 1. Characterization of biometric identification systems 
Biometric identification systems incorporate the use of physiological traits 
such as fingerprints, finger geometry, palm prints, hand geometry, hand 
topography, hand and wrist vein patterns, retina and iris characteristics, face, facial 
features, and variations in facial temperature. They also encompass the use of 
behavioral traits, such as signature, voice, keystroke, and pointing patterns. Figure 
1 depicts this characterization of biometric identification systems. 
A physiological trait is a relatively stable physical feature that is unalterable 
without trauma to the individual. By contrast, a behavioral trait, has some 
physiological basis, but also reflects a person's psychological makeup [Miller, 
1994]. The differences between physiological and behavioral methods are: 
a. The degree of intrapersonal variation is smaller in a physiological trait than 
in a behavioral one. 
b.Behavioral biometrics are influenced by both controllable and unintentional 
psychological factors, i.e., emotions, colds, fatigue, or stress. 
c. Behavioral biometric devices are often smaller, cheaper, and more friendly 
than machines that measure physiological traits. 
Face-based biometric systems. Although biometric identification systems 
based on the human face are still in their infancy, they have a very important and 
indispensable role in the field of automatic biometric identification systems. This 
role is based on the fact that face-based biometric identification systems have 

---

351 
many features that other biometric identification systems lack from both the 
ergonomical and practical points of view. 
Many biometric identification systems, if not very expensive, can at first be 
intimidating to users. For example, fingerprint and hand feature based systems may 
be associated with criminal bookings. Similarly, due to the inherent need to protect 
the eyes, some people will feel uncomfortable with the idea of having a laser 
directed on their retina every time they want to make a financial transaction [Kim, 
1995]. According to Miller (1994), both retina scan and iris pattern based systems 
involve contact and people do not want to put their eyes close to the device as is 
necessary. Face-based biometric systems, however, are less intrusive, as the user 
does not have to place a finger or hand on a reading device or place an eye near a 
scanner. Therefore, face-based biometric systems are considered as non-contact 
systems that avoid the criminal stigma that other systems might have and hence 
cause no user resistance. When practical issues are taken into consideration, 
signature-based biometric systems would be acceptable to people of all ages and 
social groups who know how to sign. For those who do not know how to sign, 
however, the technique is worthless. The practical limitations of signature and 
voice biometric identification systems and other physiological (contact) based 
systems become clear if we consider an application where the system is to 
periodically check that a computer user is an authorized one. Each time a 
periodical check (or verification) is conducted, the system has the user sign, say 
something, get closer to an infra-red device, or place his hand/finger on the reading 
device. While the keystroke/pointing pattern verification systems are suitable for 
this particular application they are not suitable for other access control 
applications. Conversely, a face-based biometric system can be used by users who 
do not know how to sign and is very practical in most identification applications 
mentioned thus far. Additionally, humans use the face as a primary method of 
recognition. NeuroMetric Vision Systems Inc., Pompano Beach, Florida, 
introduced a system that employs the human face for identification and access 
control. According to Miller (1994), the system uses an IBM-compatible personal 
computer; a frame grabber; a custom digital processing card that locates the face, 
scales and rotates it if necessary, compensates for lighting differences, and 
performs mathematical transformations to reduce the face to a set of floating point 
feature vectors; and a neural network for recognition. The system is 
currently being evaluated by Sandia National Laboratory.
The face-based biometric access control system implemented here has 
successfully integrated three identity-verification techniques and hence provided 
more secure access control against impostors. The research objectives were 
fulfilled by developing this system, which requires only a brief time for training or 
recognition (approximately three seconds on a Pentium 90 for each) because it 
employs computationally economical image processing. In addition to successfully 
implementing a face-based biometric system, the major contribution of this 
research is the integration of more than one identity verification technique in the 
developed system, which allows utilizing varying levels of security. The system is 

---

352 
suitable for office or laboratory use and the different security levels that it provides 
makes it more attractive for other applications. 
References and bibliography 
1. Allinson, N. M. and A. W. Ellis, "Face Recognition: Combining Cognitive Psychology 
and Image Engineering," Electronics and Communication Engineering Journal, Vol. 4, Iss. 5, pp. 
291-300, October 1992. 
2. Altaf, Usamah, and Cihan H. Dagli, "Face recognition Using the HAVNET Neural 
Network," in Procedings o f SPIE's Conference on Applications and Science o f 
Artificial Neural Networks, Steven K. Rogers and Dennis W. Ruck (eds.), SPIE, 
Bellingham, WA, Vol. 2492, No. II, pp. 873-883, 1995. 
3. Altaf, U., Klinkenberg R., and Dagli, C., "Automatic Face Recognition: Fuzzy 
Classification versus Neural Networks," to appear in The International Journal of 
Microcomputer Applications. 
4. Anthes, Gary H., "A Picture's Worth a Thousand Passwords," Computerworld, Vol. 29, 
No. 22, p. 66, May 1995.