At the forefront of Artificial Intelligence

Biometrics: An Overview

Biometrics is an exciting new field that has been brought into the spotlight recently due to the September 11th attacks on the United States. What exactly is biometrics? What types of biometric identification are there? How available are they?

What is "Biometrics"?

Biometrics is a modern technological field that focuses on identifying an individual through his or her unique physical traits. Biometrics spans various fields such as artificial intelligence and biology as well as various hardware-related fields. Typical methods of biometric identification include: fingerprint, voice, iris, retina, hand, facial and signature (signature isn't a physical trait but is often grouped under biometrics).

Fingerprint and Hand

Fingerprinting (also called finger scanning due to the negative connotations that 'fingerprinting' implies) is one of the most common forms of biometric identification. Fingerprinting basically uses a fingerprint's minutae - the ridges, bifurications, islands and other traits that were classified back at the end of the 1800s. The fingerprint image is collected by the software, then pre-processed by thresholding the image and simplifying the features using standard image processing techniques.

The process after this is up to the particular vendor but most will analyze the ridges and extrapolate the necessary data from that. Most fingerprint data takes up around 250Kb, although many vendors will encrypt this data for additional security.

Fingerprinting is a widely used and accepted technology with very good accuracy rates. Fingerprinting has more vendors than any other area of biometrics, such as SecuGen and BioLink USA. Also, see Generation5's review of the SecuGen EyeD Optical Mouse.

Hand technology is similar but surprisingly uses a smaller amount information. Hand technology is also suspectible to injuries to hand (fairly common) and can be fairly expensive to install. Hand recognition is dominated by one vendor, Recognition Systems.

Voice

Voice recognition is very different to speech recognition. Speech recognition detect words and sentences from an incoming audio signal whereas voice recognition tries to detect the speaker given a voice sample. Our voiceprint is determined by many different factors: size of the vocal cavities (throat, oral, nasal) as well as the characteristics of the vocal chords themselves. Your voice is further modified by the way you speak - the way your mouth, lips, tongue, jaw and teeth move (these are called articulators). Therefore, the chances of two people have the same vocal characteristics are minute.

Computers can now quickly and efficiently extract data from the voice that is the result of these characteristics - resulting in relatively accurate (about 0.3% false acceptance rate, this is relatively accurate for biometrics!) and fast identification. For more information see Generation5's review of SAFLink SaftyLatch.

Iris and Retinal

Iris recognition is the epitome of biometric identification - the entire planet could be enrolled into an iris database and there would still be a minute chance of false identification (FAR) or rejection (FRR). The iris recognition algorithms are all based upon research and patents held by Dr. John Daugman. The method in which features are extracted is incredibly complicated but in short a specialized monochrome camera takes a picture of the eye. Software then locates the iris, taking account of the eyelids as well as light reflection and eye moisture. Segments of the iris are then mapped on to hundreds of vectors, creating an IrisCode. The IrisCode is so specialized that the chances of two irises having the same code are 1 in 10⁵²!

Retinal scanning also provides a similar level of accuracy, although there is a social stigma associated with this more intrusive method. Retinal scanning uses a device to read the back of the users eye (their retina) by having them look very closely at a rotated green light, while the system reads in more than 400 different points. This gives highly accurate results with a miniscule FAR/FRR.

Facial

Facial biometrics use various features of the face to recognize or verify a user. Facial recognition is a 1-to-many mapping whereas facial verification is a more simple 1-to-1 mapping. There are four primary facial recognition techniques: eigenfaces, feature analysis, neural networking and automatic face processing. The neural network approach, for example, uses the template face and runs it through a neural network that tries to identify it with its database (or verifies it against the template the user is enrolled to). If a false acceptance or rejection occurs then the neural network modifies its weights to help improve recogntion at a later date. Another interesting approach is MIT's Eigenfaces. Basically, a database holds a large number of template faces (typically between 60-120). When your face is read in, it is mapped on to the template faces so that the program can mix the template faces together to actually artificially reproduce your face! This is then used later for recognition and verification purposes. Eigenface techniques have to be applied in fairly controlled situations like an ID photo not grainy security camera footage.

Conclusion

Biometrics is a very interesting and exciting field that has be growing exponentially in recent years (especially 2001). The wide variety of physically unique traits our bodies give us will soon allow us to live in a very secure password-less world. For more information I thoroughly recommend the International Biometric Group website - use the links at the very top of the page to find more about the individual technologies, vendors and other weblinks.

Submitted: 15/01/2002

Article Toolbar

BibTeX entry

External Links

- International Biometric Group
- Voiceprint Identification
- SecuGen
- SAFLINK