BIOMETRICS refers to the automatic identification of a person based on his or her physiological or behavioral characteristics like fingerprint, or iris pattern, or some aspects of behaviour like handwriting or keystroke patterns. Biometrics is being applied both to identity verification. The problem each involves is somewhat different. Verification requires the person being identified to lay claim to an identity. So the system has two choices, either accepting or rejecting the personâ„¢s claim. Recognition requires the system to look through many stored sets of characteristics and pick the one that matches the unknown individual being presented. BIOMETRIC system is essentially a pattern recognition system, which makes a personal identification by determining the authenticity of a specific physiological or behavioral characteristics possessed by the user.
Biometrics is a rapidly evolving technology, which is being used in forensics Such as criminal identification and prison security, and has the potential to be used in a large range of civilian application areas. Biometrics can be used transactions conducted via telephone and Internet (electronic commerce and electronic banking. In automobiles, biometrics can replace keys with key-less entry devices
BIOMETRICS refers to the automatic identification of a person based on his physiological / behavioral characteristics. This method of identification is preferred for various reasons;the person to be identified is required to be physically present at the point of identification; identification based on biometric techniques obviates the need to remember a password or carry a token. With the increased use of computers or vehicles of information technology, it is necessary to restrict access to sensitive or personal data. By replacing PINs, biometric techniques can potentially prevent unauthorized access to fraudulent use of ATMs, cellular phones, smart cards, desktop PCs, workstations, and computer networks. PINs and passwords may be forgotten, and token based methods of identification like passports and driverâ„¢s licenses may be forged, stolen, or lost .Thus biometric systems of identification are enjoying a renewed interest. Various types of biometric systems are being used for realâ€œtime identification ; the most popular are based on face recognition and fingerprint matching. However there are other biometric systems that utilize iris and retinal scan, speech, facial thermo grams, and hand geometry.
A biometric system is essentially a pattern recognition system, which makes a personal identification by determining the authenticity of a specific physiological or behavioral characteristics possessed by the user. An important issue in designing a practical system is to determine how an individual is identified. Depending on the context, a biometric system can be either a verification (authentication) system or an identification system. There are two different ways to resolve a personâ„¢s identity : Verification and Identification. Verification ( Am I whom I claim I am ?) involves confirming or denying a personâ„¢s claimed identity. In Identification one has to establish a personâ„¢s identity (whom am I?). Each one of these approaches has its own complexities and could probably be solved best by a certain biometric system.
Biometrics is rapidly evolving technology, which is being used in forensics such as criminal identification and prison security, and has the potential to be used in a large range of civilian application areas . Biometrics can be used transactions conducted via telephone and Internet (electronic commerce and electronic banking) . In automobiles, biometrics can replace keys with key -less entry devices.
2. ORIGIN OF BIOMETRICS
Biometrics dates back to the ancient Egyptians, who measured people to identity them. But automated devices appeared within living memory. One of the first commercial devices introduced less than 30 years ago. The system is called the indentimat . The machine measured finger length and installed in a time keeping system. Biometrics is also catching on computer and communication system as well as automated teller machines (ATMâ„¢s).
Biometrics devices have three primary components. One is an automated mechanism that scans and captures a digital / analog image of a living personal characteristics. Another handles compression, processing, storage and comparison of image with the stored data . The third interfaces with application systems. These pieces may be configured to suit different situations . A common issue is where the stored image resides:on a card, presented by the person being verified or at a host computer.
Recognition occurs when an individualâ„¢s image is matched with one of a group of stored images . This is the way the human brain performs most day to day identifications. For the brain this is a relatively quick and efficient process, where as for computers to recognise that a living image matches one of many it has stored, the job can be time consuming and costly.
3. TYPOLOGY OF BIOMETRICS
Biometrics encompasses both physiological and behavioural characteristics. This is illustrated in Figure 1. A physiological characteristic is a relatively stable physical feature such as finger print, hand silhouette , iris pattern or facial features. These factors are basically unalterable with out trauma to the individual.
A behavioral tract, on the other hand, has some physiological basis, but also reflects personâ„¢s physiological makeup. The most common trait used in identification is a personâ„¢s signature. Other behaviours used include a personâ„¢s keyboard typing and speech patterns. Because of most behavioural characteristics change over time, many biometrics machine not rely on behavior. It is required to update their enrolled reference template may differ significantly from the original data, and the machine become more proficient at identifying the person. Behavioral biometrics work best with regular use.
The difference between physiological and behavioral methods is important. The degree of intrapersonal variation is smaller in physical characteristics than in a behavioral one. Developers of behaviour-based systems, therefore have a tougher job adjusting for an individualâ„¢s variability. However, machines that measure
physical characteristics tend to be larger and more expensive, and more friendly. Either technique affords a much more reliable level of identification than passwords or cards alone.
TYPOLOGY OF IDENTIFICATION METHODS
4. VARIOUS BIOMETRIC SYSTEMS
The three dimensional shape of a personâ„¢s hand has several advantages as an identification device. Scanning a hand and producing a result takes 1.2 seconds. It requires little space for data storage about 9 bytes which can fit easily magnetic strip credit cards.
Hand geometry is the grand daddy of biometrics by virtue of its 20 year old history of live application. Over this span six hand-scan products have been developed but one commercially viable product currently available, the ID3D hand key is given below. This device was developed by Recognition Systems Inc.
The user keys, in an identification code, is then positions his or her and on a plate between a set of guidance pins. Looking down upon the hand is a charge-coupled device (CCD) digital camera, which with the help of mirror captures the side and top view of the hand simultaneously.
The black and white digital image is analysed by software running on a built in HD 64180 microprocessor. ( This a Z-80 base chip ) to extract identifying characteristics from the hand picture. The software compares those features to captured when the user was enrolled in the system, and signals the result-match or no match. Analysis is based on the measurement and comparison of geometric. The magnification factor of the camera is known and is calibrated for pixels per inch of real distance. Then the dimensions of parts of the hand, such as finger length, width and area are measured, adjusted according to calibration marks on the platen and used to determine the identifying geometric of the hand.
A strong correlation exists between the dimension of the hand. For example if the little finger is long, the index finger will most likely also be along. Some 400 hands were measured to determine these interrelationships, and the results are integrated into the system as a set of matrices are applied to measured geometric to produce the 9 byte identity feature vector that is stored in the system during enrolment, with this amount of data compression, the current 4.5 kg unit with single printed circuit board can store 2000 identities.
Enrolment involves taking three hands reading and averaging the resulting vectors. Users can enrol themselves with minimal help. When used for identification the 9-byte vector is compared to the stored vector and score based on the scalar difference is stored. Low scores indicate a small difference, high scores mean a poor match. The recognition systems product fine-tunes the reference vector a small increment at a time, in case the original template was made under less than perfect conditions.
There are so many other systems for hand recognition. One was an effort by SRI international, to take pictures of unconstrained hands help in free space. This system was introduced in 1985. Biometrics Inc., Tokyoâ„¢s Toshiba Corp. Identification corp. etc are some companies which developed biometrics systems.
4.2 FINGER PRINT
Perhaps most of the work in biometrics identification has gone into the fingerprint For general security and computer access control application fingerprints are gaining popularity.
The fingerprintâ„¢s stability and uniqueness is well established. Based upon a century of examination, it is estimated that the change of two people, including twins, having the same print is less than one a billion. In verifying a print, many devices on the market analyze the position of details called minutiae such as the endpoints and junctions of print ridges. These devices assign locations to the minutiae using x, y, and directional variables. Some devices also count the number of ridges between minutiae to form the reference template. Several companies claim to be developing templates of under 100 bytes. Other machine approach the finger as an image processing problem and applying custom very large scale integrated chips,neural networks, fuzzy logic and other technologies to the matching problem.
The fingerprint recognition technology was developed for some 12 years before Being matched in 1983 by Identix Inc.
The Identix system uses a compact terminal that incorporates light and CCD image sensors to take high-resolution picture of a fingerprint. It based on 68000 CPU with additional custom chips, but can also be configured as a peripheral for an IBM PC. It can operate as a standalone system or as part of a network.
To enrol a user is assigned a personal identification number and then puts a single finger on the glass or Plexiglas plate for scanning by a CCD image sensor. The 250-KB image is digitalized and analyzed, and the result is approximately 1-KB mathematical characterization of the fingerprint. This takes about 30 seconds. Identity verifications take less than 1 second . The equipment generally gives the user three attempts for acceptance or finds rejection. With the first attempt the false rejection is around 2-3 percent and false acceptance is less than 0.0001 per cent. Each standalone unit cab stores 48 fingerprint templates which may be expanded to 846 by installing an additional memory package.
Fingerprints have overcome the stigma of their use in law enforcement and military applications. Finger print recognition is appropriate for many applications and is
familiar idea to most people even if only from crime dramas on television. It is non-intrusive, user friendly and relatively inexpensive.
Biometrics developers have also not lost sight of fact that humans use the face as their primary method of telling whoâ„¢s who. More than a dozen effort to develop automated facial verification or recognition systems use approaches ranging from pattern recognition based on neural networks to infrared scans of Ëœhot spotsâ„¢ on the face.
Using the whole face for automatic identification is a complex task because its appearance is constantly changing. Variations in facial expressions, hair styles and facial hair, head position, camera scale and lighting create image that are usually different from the image captured on a film or videotape earlier. The application of advanced image processing techniques and the use of neural networks for classifying the images, however, has made the job possible.
Artificial neural networks are massively connected parallel networks of simple computing elements. Their design mimics the organization and performance of biological neural networks in the nervous system and the brain. They can learn and adapt and be taught to recognize patterns both static and dynamic. Also their interconnected parallel structure allows for a degree of fault tolerance as individual computing elements become inoperative. Neural networks are being used for pattern recognition function approximation, time series analysis and disk control.
There is only one system available on the market today. The system is developed by Neuro Metric Vision system Inc. this can recognize faces with a few constraints as possible, accommodating a range of camera scales and lighting environments, along with changes in expression and facial hair and in head positions. The work sprang from the realisation that such techniques as facial image comparisons, measurement of key facial structure and the analysis of facial geometry could be used in face recognition system. Any of these approaches might employ rule-based logic or a neural network for the image classification process.
The Nuerometric system operates on an IBM-compatible 386 or 486 personal computer with a maths co-processor, a digital signal processing card and a frame grabber card to convert raster scan frames from an attached camera in to pixel representations. The system can capture images from black and white video cameras or vide recorders in real time.
Software running on the DSP card locates the face in the video frame, scales and rotates if necessary, compensating for lighting differences and performs mathematical transformations to reduce the face to a set of floating point feature vectors. The feature vector set is input to the neural network trained to respond by matching it to one of the trained images in as little as 1 seconds.
The systemâ„¢s rejection level can be tuned by specifying the different signal to noise ratios for the match â€œ a high ratio to specify a precise match, and a lower one to allow more facial variation. In a tightly controlled environment, for example, the system could set up to recognise a person only when looking at the camera with same expression he or she had when initially enrolled in the system.
To enrol someone in the Neuro Metric system, the face is captured, the feature vectors extracted, and the neural network is trained on the features. Grayscale facial images may be presented from live video or photographs via videodisk. The neural network is repeatedly trained until it learns all the faces and consistently identifies every image. The system uses neural network clusters of 100-200 faces to build its face recognition database. If multiple clusters are required they can be accessed sequentially or hierarchically. When faces are added to or detected from the database, only the affected clusters must be retrained, which takes 3-5 minutes.
The other method of identification involves the eye. Two types of eye identification are possible, scanning the blood vessel pattern on the retina and examining the pattern of the structure of the iris. Now we can look through a detailed description of each type below.
4.4 1 RETINA
Retina scans, in which a weak infrared light is directed through the pupil to the back of the eye, have been commercially available since 1985. The retinal pattern is reflected back to a charge-coupled device (CCD) Camera, which captures the unique pattern and represents it in less than 35 bytes of information. Retina scans are one of the best biometrics performers on the market, with low false reject rates and nearly 0 present false accept rate. The technology also offers small data templates provides quick identity confirmations, and handles well the job of recognizing individuals in a database of under 500 people. The toughest hurdle for retinal scan technology is user resistance. People donâ„¢t want to put their eye as close to the device as necessary. Only one company, Eyedentyfy Inc., produces retinal scan products.
4.4 2 IRIS
Once it was the whites of their eyes that counted. Retinal pattern recognition has been tried but found uncomfortable because the individual must touch or remain very close to a retinal scanner. Now the iris is the focus of a relatively new biometrics means of identification. Standard monochrome video or photographic technology in combination with robust software and standard video imaging techniques can accept or reject an iris at distance of 30-45 cm.
A device that examines the human iris is being developed by Iriscan Inc. The techniqueâ„¢s big advantage over retinal scans is that it does not require the user to move close to the device and focus on a target because the iris pattern is on the eyeâ„¢s surface. In fact the video image of an eye can be taken at distance of a metre or so, and the user need not interact with device at all.
The technology being implemented by Iriscan Inc., is based on principles developed and planted by ophthalmologists Leonard Flom and Aran Safir and on mathematical algorithms developed by John Daugman. In their practice, Flom and Safir observed that every iris had highly detailed and unique texture that remains stable over decades of life. This part of the eye is one of the most striking features of the face. It is easily visible from yards away a s a coloured disk, behind the clear protective window of the cornea, surrounded by the white tissue of the eye. Observable features include contraction furrows striations, pits, collagenons fibres, filaments, crypts, serpentine, vasculature, rings and freckles. The structure of iris is unique, as in fingerprint, but it boasts more than six times as many distinctly different characteristics as the finger print. This part of the eye, moreover cannot surgically modified without damage to vision. It is produced from damage or internal changes by the cornea and it responds to light, a natural test against artifice.
Another biometrics approach that is attractive because of its acceptability to users is voice verification. All the systems used in analyzing the voice are rooted in more broadly based speech processing technology. Currently, voice verification is being used in access control for medium security areas or for situations involving many people as in offices and lab. There are two approaches to voice verification. One is using dedicated hardware and software at the point of access .The second approach is using personal computer host configurations that drives a network over regular phone lines.
One of the latest implementation of the technology is the recently demonstrated AT&T Smart Card used in an automatic teller system. The AT&T prototype stores an individualâ„¢s voice pattern on a memory card, the size of a credit card. In brief, someone opening an account at a bank has to speak a selected two or three-syllable word eight items. The word can be chosen by the user and belong to any language or dialect.
Another approach being as an alternative to the algorithms discussed is based on Hidden Markov Models, which consider the probability of state changes and allow the system to predict what the speaker is trying to say. This capability would be crucial for speaker independent recognition. Storing voice templates on a card and receiving and processing voice information at a local device, such as ATM, eliminated variations due to telephone connection and types of telephones used.
4.5.1 SPEAKER VERIFICATION
The speaker- specific characteristics of speech are due to differences in physiological and behavioral aspects of the speech production system in humans. The main physiological aspect of the human speech production system is the vocal tract shape. The vocal tract is generally considered as the speech production organ above the vocal folds, which consists of the following: (a) laryngeal pharynx ( beneath the epiglottis), (b) oral pharynx ( behind the tongue, between the epiglottis and velum ), ( c) oral cavity ( forward of the velum and bounded by the lips, tongue, and palate ), (d) nasal pharynx ( above the velum, rear end of nasal cavity ), and (e) nasal cavity (above the palate and extending from the pharynx to the nostrils ). The shaded area in figure 4 depicts the vocal tract.
The vocal tract modifies the spectral content of an acoustic wave as it passes through it, thereby producing speech. Hence, it is common in speaker verification systems to make use of features derived only from the vocal tract. In order to characterize the features of the vocal tract, the human speech production mechanism is represented as a discrete-time system of the form depicted in figure 5.
The acoustic wave is produced when the airflow from the lungs is carried by the trachea through the vocal folds. The source of excitation can be characterized as phonation, whispering, friction, compression, vibration, or a combination of these. Phonated excitation occurs when the airflow is modulated by the vocal folds. Whispered excitation is produced by airflow rushing through a small triangular opening between the arytenoids cartilage at the rear of the nearly closed vocal folds. Friction excitation is produced by constrictions in the vocal tract. Compression excitation results from releasing a completely closed and pressurized vocal tract. Vibration excitation is caused by air being forced through a closure other than the vocal folds, especially at the tongue. Speech produced by phonated excitation is called voiced, that produced by phonated excitation plus friction is called mixed voiced, and that produced by other types of excitation is called unvoiced.
It is possible to represent the vocal-tract in a parametric form as the transfer function H (z). In order to estimate the parameters of H (z) from the observed speech waveform, it is necessary to assume some form for H (z) . Ideally, the transfer function should contain poles as well as zeros. However, if only the voiced regions of speech are used then an all-pole model for H (z) is sufficient. Furthermore, linear prediction analysis can be used to efficiently estimate the parameters of an all-pole model. Finally, it can also be noted that the all-pole model is the minimum-phase part of the true model and has an identical magnitude spectra, which contains the bulk of the speaker-dependent information.
4.6 MULTI BIOMETRICS
4.6.1 Integrating Faces and Fingerprints for Personal Identification
An automatic personal identification system based on fingerprints or faces is often not able to meet the system performance requirements. Face recognition is fast but not reliable while fingerprint verification is reliable but inefficient in database retrieval. A prototype biometric system is developed which integrates faces and fingerprints. The system overcomes the limitations of face recognition systems as well as fingerprint verification systems. The integrated prototype system operates in the identification mode with an admissible response time. The identity established by the system is more reliable than the identity established by a face recognition system. In addition, the proposed decision fusion schema enables performance improvement by integrating multiple cues with different confidence measures. experimental results demonstrate that our system performs very well. It meets the response time as well as the accuracy requirements.
4.6.2 A Multimodal Biometric System Using Fingerprint, Face
A biometric system which relies only on a single biometric identifier in making a personal identifications often not able to meet the desired performance requirements. Identification based on multiple biometrics represents on emerging trend. A multimodal biometric system is introduced (figure given below ), which integrates face recognition, fingerprint verification, and speaker verification in making a personal identification.
This system takes advantage of the capabilities of each individual biometric. It can be used to overcome some of the limitations of a single biometrics. Preliminary experimental results demonstrate that the identity established by such an integrated system is more reliable than the identity established by a face recognition system, a fingerprint verification system and a speaker verification system.
A range of biometric systems are in developments or on the market because no one system meets all needs. The trade off in developing these systems involve component cost, reliability, discomfort in using a device, the amount of data needed and other factors. But the application of advanced digital techniques has made the job possible. Further experiments are going all over the world. In India also there is a great progress in this field. So we can expect that in the near future itself, the biometric systems will become the main part in identification purposes.
2. BIOMEDICAL INSTRUMENTATION W.H. CROWELL
3. PENSTROKES AUGUST 2002
I express my sincere thanks to Prof. M.N Agnisarman Namboothiri (Head of the Department, Computer Science and Engineering, MESCE), Mr. Zainul Abid (Staff incharge) for their kind co-operation for presenting the seminars.
I also extend my sincere thanks to all other members of the faculty of Computer Science and Engineering Department and my friends for their co-operation and encouragement.
Chapter Title page
1 INTRODUCTION 1
2 ORIGIN OF BIOMETRICS 3
3 TYPOLOGY OF BIOMETRICS 4
4 VARIOUS BIOMETRIC SYSTEMS 6
4.1 HAND 6
4.2 FINGERPRINT 8
4.3 FACE 11
4.4 EYE 13
4.5 SPEECH 15
4.6 MULTI BIOMETRICS 19
5 CONCLUSION 22
6 REFERENCES 23