Most Biometric systems are used as a means of authentication when a primary means of identification, such as a card, is presented. Other more sophisticated systems are employed for primary identification, requiring no cards, passwords, or PINs. These "automated positive identification" Biometric systems prevent multiple enrollments by capturing, recording, and comparing an individual’s physical trait against an entire database as opposed to checking one record for a match. The cost and complexity of these types of Biometric systems have tended to limit their use to security applications, but as cost comes down, and processing power continues to increase, these systems will see more general use.

For Biometric identification, selection of a stable physical characteristic is key. Stable characteristics include the user’s hand silhouette, a facial feature, iris pattern, a blood vessel pattern on the retina or hand, and of course, a fingerprint. Individual behaviors may also be used for biometric identification. Behavioral identification may be achieved by analyzing signature dynamics e.g. how one types on a keyboard, or how one speaks (voice patterns). For example, signature dynamics differentiate the parts of the signature that are habitual from those that vary every time you sign your name. Because behavioral characteristics vary over time, behavioral-based equipment may update users’ enrolled Biometric reference templates each time they access the system. With each use, the machine becomes increasingly proficient at identifying an individual.

Performance of Biometric systems is measured by their identifying power, which is calculated using false-rejection and false-acceptance rates. Biometric identification systems allow users to set the desired balance of false-rejection and false-acceptance. If this tolerance is tightened to make it harder for imposters to beat the system, it is also harder for authorized people to access it. Biometric experts state that thorough user training is the best way to reduce false rejections. Knowing and optimizing a system’s identifying power, and making sure it is acceptable for your application and in your industry, are critical for system success. For example, adoption of automated signature verification for credit card and check applications has been slow because the financial community demands very low false rejection rates.

• Provides the basis for dispensing with conventional personal identification (or verification) methods based upon passwords, tokens, ID cards, and personal identification numbers (PINS).
• Performance determined by a number of factors requiring performance ratings and capabilities to be expressed for products on offer.
• Range of techniques and products emerging to satisfy a wide variety of application needs.
• Costs reducing as more applications are identified and accommodated by available products.
• Performance ratings improving as systems are further developed and experience is gained in their use, including the use of double or more biometrics where two or more biometric feature sets are used for identification (or verification) purposes.
• Biometric templates may be used in conjunction with card-based data carriers (magnetic stripe, smart cards, multi-row bar codes and matrix code data carriers).

Biometrics Using Fingerprint
The best known biometric systems are those that identify persons using their fingerprints. The chance of two people having the same print is less than one in one billion. False-rejection rates average 3 percent of authorized users; false-acceptance rates are less than one in one million. Automatic fingerprint identification systems (AFIS) have traditionally been used by law enforcement agencies. Use is now expanding into social services applications to identify those who claim welfare benefits, to monitor prisoner movement, to confirm health spa membership, at border crossings, and even at banking kiosks. AFIS systems are also gaining in popularity for security and access control applications. Fingerprint identification, coupled with time and attendance software, prevents employees from "buddy punching."

A typical AFIS system requires a user to place a finger on the machine for as little as one-half second to two seconds. Many devices analyze the position of the end points and junctions of print ridges (minutiae) of the fingerprint. Others count the number of ridges between points, while some approach the fingerprint from an image processing perspective. For storage, a fingerprint requires a data template that ranges from several hundred bytes to more than 1,000 bytes depending on the approach. Several companies with devices in development claim they will have templates under 100 bytes.

Some systems combine smart card technology and live fingerprint scanning. A digital template of the user’s fingerprint (no ink involved!) may be captured and stored with credit card information on a smart card. Smart card readers can be integrated into point-of-sale terminals. When the card is inserted into the terminal, the cardholder is prompted to place a finger on the integrated fingerprint scanner to determine if the live scan and the information on the card match. Major credit card companies have long-term plans to incorporate this technology for use when smart cards become more common and when smaller, less expensive hardware is widely available.

Biometrics Using Hand Geometry
Biometric systems that employ hand geometry have been in use at application sites for 20 years and are deployed successfully at thousands of locations, including welfare agencies and other government bodies, sperm banks, daycare centers and immigration facilities. Even a large university has used hand geometry for some time to ensure the integrity of its "all you can eat" meal program. A typical hand geometry identification system looks at both the top and side view of a person’s hand using a video camera and compression algorithms. The reference template is an economical 10 bytes or less. Other devices look at features such as the pattern of blood vessels in the hand.

Biometrics Using Retinal Scan
The blood vessel patterns of the retina and the pattern of flecks on the iris both offer unique methods of identification. These methods are presently used for high security access control at military and bank facilities. Retinal recognition is said to provide the most stable means of biometric identification over time. Orientation problems are minimized because the eye naturally aligns itself as it focuses on an illuminated target. However, comparisons of template records can take upwards of 10 seconds, depending on the size of your database. Initial enrollment requires 15 to 20 seconds per record.

Iris scanning does not require the person to interact with a device; a video image of the eye can be taken from one foot away. This has obvious benefits in applications like the one to positively identify prisoners described at the end of this section. The user’s iris pattern is reflected back to the camera, which captures the unique pattern and stores it using less than 35 bytes of information. Like iris scanning, facial feature identification systems can capture images from a distance (several meters) using video equipment. As in other more complex systems, the challenge is achieving high levels of performance as the size of the database increases. The potential of these systems is generating much interest. Increased development efforts are needed in the areas of multimedia video technology and the complex software that facial identification requires.

Biometrics Using Voice Patterns
Speech patterns encompass both physiological and behavioral factors, and voice identification devices focus on different characteristics than does the human ear. In other words, an imposter may be able to imitate someone’s voice extremely well, but will not fool a voice identification system. Voice pattern analysis systems may be set up with dedicated hardware and software at the access point, or users may achieve access by phone. One current application for voice verification systems is to monitor computer use. Biometric identification through voice pattern analysis is one of the most acceptable methods to users.

Acknowledgement: Some of the information on AIDC pages is based on the information in AIMGlobal's website. We would like to thank AIMGlobal for this.