Biometric features
When we are speaking about biometrics we are referring to the automatic identification of a person based on some of his/her physiological or behavioural characteristics.
It is possible to identify a number of characteristics, which are unique and can identify the person. Those characteristics are divided into two categories; the physiological and the behavioural, as presented in Figure 1.
On one hand, within physiological group, we can consider the characteristics of the face, fingerprint, hand geometry, and eye, namely iris and the retina of eye.
On the other hand, within behavioural group, we are considering the signature, which can be dynamic or static, the voice and the writing speed.
Figure 1 Biometric features classification
Goals of biometric-based systems
A biometric system is essentially a pattern recognition system, which makes a personal identification by determining the authenticity of a specific physiological or behavioural characteristic of the user.
A biometric system can be either a verification (authentication) system or an identification system. This means that using biometric features it is possible to identify a person (Who I am?) or verify the identity of a person (Am I who I claim I am?).
In the access control domain, it is common to use biometric-based systems as a verification system.
Most of these systems generally use a kind of camera for image acquisition, so they rely on image processing. In the case of voice recognition, they use some sort of sound acquisition equipment or sound recorder and are based in sound processing (digital signal processing).
Common methods in use when implementing biometric-based identification or verification systems use neural networks, genetic algorithms or/and fuzzy logic technology. The fact that the verification process is not as simple as a comparison with a PIN number or an ID number implies that the systems is not error-free, in the sense that sometimes we accept an unauthorized user or reject an authorized user. These faults measure the reliability of the systems. In those terms we can define two factors, one is the false rejection rate and the other is the false acceptance rate.
“The false rejection rate, or FRR, is the measure of the likelihood that the biometric security system will incorrectly reject an access attempt by an authorized user. A system’s FRR typically is stated as the ratio of the number of false rejections divided by the number of identification attempts.”
“The false acceptance rate, or FAR, is the measure of the likelihood that the biometric security system will incorrectly accept an access attempt by an unauthorized user. A system’s FAR typically is stated as the ratio of the number of false acceptances divided by the number of identification attempts.”
These two factors are dependent one on the other, in the sense that if we decrease the FAR, then FRR will increase. The ideal solution is to have both equal to zero, but it is not possible. So, according with the specific techniques used by the system, the goal is to choose the best parameters guaranteeing both low rates.
As examples, typical values for FAR for hand geometry (shape) are low (around 0.2%), while for retina are also very low, close to 0. On the other hand, FRR for hand geometry (shape) are low (around 0.2%), while for retina are high (around 1.5%). Figure 2 presents indicative figures for several common techniques.
Figure 2 Comparative False Acceptance Rates and False Rejection Rates for several common techniques
Face
Maybe the most frequently used method for identification among humans is the face recognition. We are using it every day when looking at someone. Although, it has to be stressed that using all regions of the face within an automatic recognition system is a complex problem.
Although the face is considered a physiological feature, it has changes depending on specific behaviour such as some emotions, or on hair and beard. Usually, for image grabbing, a camera such as a web cam is used.
The quality of the image depends, among others, on illumination, zoom, and distance of the camera. Those conditions and also the position of the head can difficult the identification/verification process. The captured image is compared with the images on the database.
Recently, after 2001, this method has been used more frequently on airports, relying on usage of neuronal networks for image classification.
Hand
Using hand geometry as biometric feature involves analyzing and measurement of the hand’s shape and of the fingers’ length. The image acquisition system is composed by a light source, a CCD camera (charge-coupled device) and a flat surface (with five pegs on it). Some systems also include a single mirror for capture an image not only from top view, but also from lateral view.
The user places his/her hand – palm facing downwards – on the flat surface of the device. The five pegs serve as control points for an appropriate placement of the right hand of the user. A black and white image is acquired and analyzed. Feature extraction commonly involves computing the widths and length of fingers at several locations, among others.
Figure 3. Hand-based acquisition set-up
Using hand geometry for identification/verification has some advantages:
- Is rapid;
- Needs only few memory, typically less than 10 bytes;
- Easy to be used;
- User friendly.
Alternatively to hand geometry (shape), some works refer hand veins usage instead, which can improve the performance of the system.
Most of commercial available systems are for identity verification, which means that prior to perform image acquisition, feature extraction and classification/identification/verification, the user inputs a personal code (PIN), using a dedicated keyboard.
Eye
There are two identification methods involving eyes’ feature extraction and processing: examine the veins on retina and the other one is examine the structure of iris.
In case of the retina, a weak infrared light is directed through the pupil to behind of the eye. This light is reflected on the retina and is captured by a CCD camera. Typically, associated features can be stored in less than 35 bytes.
Main advantages of this method are the following:
- Is rapid;
- Based on small images;
- Reliable not only for verification but also for identification.
Even though, it is not so common to use this method because it is too intrusive, and, in average, people don’t like it. Due to infrared light propagation characteristics, it is necessary to put the eye too close to the reader, even almost is necessary to touch it. Many people refuse to use it.
Examination of the iris is more suitable, because the fact that the iris is on the surface of the eye it is not required to be too much close to the reader. In fact it is possible to capture an image from a considered distance, even almost half meter. Patterns in the iris contain more than 200 unique spots that differentiate one eye from another. The identification or verification is made by using a series of complex computer image algorithms to matching the captured image with the images on the database.
Fingerprint
Fingerprint recognition is the oldest method of biometric identification. Its history is retracing at least 6000 BC. The Assyrians, the Babylonians, the Chinese and the Japanese already used the fingerprint as a personal code. Since 1897 for criminal identification, the fingerprint identification technique has been used, named as dactyloscopy.
The fingerprint is composed by ridges (lines across fingerprints) and valleys (spaces between ridges). The pattern of fingerprint is unique for each individual and it is immutable.
Fingerprint matching techniques can be placed into two categories:
- Minutiae-based, and
- Correlation based.
The first approach is based on analysis of ridge bifurcations and endings. However, there are some difficulties. For example, when the fingerprint is of low quality it is difficult to extract these minutiae points accurately; also, this method fails to take into account the global pattern of ridges and valleys.
The second method is able to overcome some of these difficulties, represents a more macroscopic approach, considering the flow of ridges in terms of arches, loops and whorls, for example. However it has also some difficulties, such as require the precise location of a registration point and are affected by the image translation or rotation.
Voice
Using voice recognition as an identification method for access control applications can be considered as the less invasive method. This is because it is based on something that we are used to do every day: talking.
This method is composed by several stages, starting with the acquisition/record of the voice, extract some features (characteristics), such as spectral analysis, tone, speed, and energy density, among others. This data can be storage on a card, such as a magnetic card or a chip card, which can be carried by the user.
This means that it is not necessary to have a big database associated with the system. The user can put the card into the reader, saying the known words and the system will make the comparison and conclude if the extracted characteristics are matching with the data stored in the card.
This method is user friendly, but the associated processing can be very complex. Also, it is not independent of the noise, coming from the environment.