Biometrics

Biometrics:

Biometrics refers to the automatic identification of a person based on his/her physiological or behavioral characteristics. This method of identification is preferred over current methods involving passwords and PIN numbers 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 as vehicles of information technology, it is necessary to restrict access to sensitive/personal data. By replacing PINs, biometric techniques can potentially prevent unauthorized access to or 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 driver'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 real-time 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 thermograms, 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.

Verification vs Identification:

There are two different ways to resolve a person's identity: verification and identification. Verification (Am I whom I claim I am?) involves confirming or denying a person's claimed identity. In identification, one has to establish a person's identity (Who am I?). Each one of these approaches has it's own complexities and could probably be solved best by a certain biometric system.

Applications:

Biometrics is a rapidly evolving technology which is being widely 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 to prevent unauthorized access to ATMs, cellular phones, smart cards, desktop PCs, workstations, and computer networks. It can be used during transactions conducted via telephone and internet (electronic commerce and electronic banking). In automobiles, biometrics can replace keys with key-less entry devices.

RECOGNIZING PERSONS BY THEIR IRIS PATTERNS

John Daugman, Ph.D.

University of Cambridge
The Computer Laboratory
Cambridge CB2 3QG England

Abstract:

Performance of face recognition schemes is bounded at one limit by a genotypic error rate (the birth rate of identical twins), and another limit by a phenotypic error rate (change in facial appearance over time). These set minimal False Accept and False Reject frequencies, by undermining the between-class variability in the first case, and increasing the within-class variability in the second. It would be preferable to base recognition decisions upon features which had very little genetic penetrance, yet high complexity, and stability over the lifetime of the individual. Phenotypic facial features do exist with exactly these properties. When imaged at a distance up to about arm's length, the population entropy (information density) of iris patterns is roughly 3.4 bits per square millimeter, and their complexity spans about 266 independent degrees-of-freedom. These statistics exceed significantly the degrees of randomness and complexity available in other identifying biometric patterns.

Introduction:

The central issue in pattern recognition is the relation between within-class variability and between-class variability. These are determined by the degrees of freedom spanned by the pattern classes. Ideally the within-class variability should be small and the between-class variability large, so that the classes are well separated.

In the case of encoding faces for identity, one would like different faces to generate face codes that are as different from each other as possible, while different images of the same face should ideally generate similar codes across conditions. Several recent investigations of how well this goal is achieved have studied the invariances in face coding schemes under changes in illumination, perspective angle or pose, and expression. Their results have tended to show that there is greater variability in the code for a given face across these three types of changes, than there is among the codes for different faces when these three factors are kept constant. Further and more fundamental difficulties arise because the birth rate of identical twins (about 0.86%) undermines the between-class variability (i.e. different persons with identical appearance), while the change in our facial appearance as we age increases the within-class variability. These are grave limitations for efforts to achieve high performance recognition of persons using gross facial appearance, despite the obvious desirability of non-contact, non-intrusive recognition from a distance using only a video camera. However, these desirable goals can be achieved by focusing instead on just a particular phenotypic feature within a face.

By far the most numerous and dense degrees-of-freedom (forms of variability among individuals), which are both stable over time and visible in a face, are found in the complex texture of the iris of either eye. This protected internal organ, which can be imaged adequately at distances of up to about a meter, reveals about 266 independent degrees-of-freedom of textural variation across individuals.

One way to calibrate the "information density" of the iris is by its human-population entropy per unit area. This works out to 3.4 bits per square millimeter on the iris, based upon 222,743 IrisCode comparisons recently collected by the British Telecom Research Laboratories using the algorithms for iris encoding and recognition to be described here.

The iris is a protected internal organ of the eye, behind the cornea and the aqueous humour, yet readily visible externally at a comfortable distance because these optical media in front of it are, of course, transparent. The iris is composed of elastic connective tissue, the trabecular meshwork, whose prenatal morphogenesis is completed during the 8th month of gestation. It consists of pectinate ligaments adhering into a tangled mesh revealing striations, ciliary processes, crypts, rings, furrows, a corona, sometimes freckles, vasculature, and other features. During the first year of life a blanket of chromatophore cells usually changes the colour of the iris, but the available clinical evidence indicates that the trabecular pattern itself is stable throughout the lifespan. [Clinical reference: Adler, 1965.]

Being an internal organ of the eye, the iris is immune (unlike fingerprints) to environmental influences, except for its pupillary response to light. The elastic deformations that occur with pupillary dilation and constriction are readily reversed mathematically by the algorithms for localizing the inner and outer boundaries of the iris. Pupillary motion, even in the absence of changes in illumination (termed "hippus"), and the associated elastic deformations it creates in the iris texture, provide one test against photographs, glass eyes, or other simulacra for a living iris. Other tests involve changing infrared LED light sources which should cause corresponding changes in their specular reflections from the cornea; detecting the properties of contact lens which might contain a printed fake iris pattern riding upon the spherical surface of the cornea, rather than in an internal plane within the eye; testing for the properties of living tissue under varying wavelengths of both visible and infrared illumination; and so forth.

 

 

Analysis of Iris Patterns:

The two-dimensional modulations that form iris patterns can be extracted by mathematical demodulating operations [Daugman and Downing, 1995], using complex-valued 2D Gabor wavelets. First it is necessary to localize precisely the inner and outer boundaries of the iris, and to detect and exclude eyelids if they intrude. These detection operations are accomplished by integro-differential operators. Then a doubly-dimensionless coordinate system is defined which maps the tissue in a manner that is invariant to changes in pupillary constriction and overall iris image size, and hence also invariant to camera zoom factor and distance to the eye. This coordinate system is pseudo-polar, although it does not assume concentricity of the inner and outer boundaries of the iris since the pupil is normally somewhat nasal, and inferior, to the iris. The coordinate system compensates automatically for the stretching of the iris tissue as the pupil dilates. It is illustrated graphically in Figure 2, together with a phase-demodulation IrisCode indicated in the top left as a bit stream containing 256 bytes of information.

Making Identifications by Comparison of IrisCodes:

The histogram in Figure 3 compares different eyes' IrisCodes by vector Exclusive-OR'ing them in order to detect the fraction of their bits that disagree. Since any given bit is equally likely to be set or clear, an average Hamming Distance fraction of 0.5 would be expected. The observed mean was 0.498 in comparisons between 222,743 different pairings of IrisCodes enrolled by British Telecom. The standard deviation of this distribution, 0.0306, indicates that the underlying number of degrees-of-freedom contained within such IrisCodes is 266. The solid curve shown fitted to the data is a binomial distribution with 266 degrees-of-freedom; this is the expected distribution after tossing a fair coin 266 times in a row, and tallying up the fraction of heads in each such run. The factorials which dominate the tails of such a distribution make it astronomically improbable that two different IrisCodes having these many degrees-of-freedom could accidentally disagree in much fewer than half their bits. The essence of iris recognition is the exploitation of such combinatorics (the branch of mathematics concerned with analyzing combinations of events and their associated probabilities.)

For example, the chances of two different IrisCodes disagreeing in only 25% or fewer of their bits (getting a Hamming Distance below 0.25, i.e.equivalent to the chances of getting fewer than 25% or 67 heads among 266 coin tosses) are less than one in 10-to-the-16th power (1 in 10,000,000,000,000,000). Thus the observation of a match with even such poor quality (one-quarter of the bits in the two IrisCodes were different) is extraordinarily compelling evidence of identity, because the combinatorial test for statistical independence has failed on such a colossal scale.

			Figure 3: Histogram of raw Hamming Distances between 222,743 pairs o Color f unrelated IrisCodes. The fitted curve is a binomial distribution with 266 degrees-of-freedom.

Genetically Identical Irises:

Just as the striking visual similarity of identical twins reveals the genetic penetrance of overall facial appearance, a comparison of genetically identical irises reveals that iris texture is a phenotypic feature, not a genotypic feature. A convenient source of genetically identical irises are the right and left pair from any given person. Such pairs have the same genetic relationship as the four irises of two identical twins, or indeed in the probable future, the 2N irises of N human clones. Eye colour of course has high genetic penetrance, as does the overall statistical quality of the iris texture, but the textural details are uncorrelated and independent even in genetically identical pairs. This is shown in Figure 4, comparing 648 right/left iris pairs from 324 persons.

		Figure 4: Histogram of raw Hamming Distances between IrisCodes computed from 324 pairs of genetically identical irises (648 eyes in right/left pairs).

The expected shape of this extreme-value distribution can be derived from the original binomial form, and this theoretical density function is shown in Figure 6. The areas under the left tail of this distribution

		Figure 6: Theoretical density distribution for best Hamming Distance matches between unrelated IrisCodes after multiple relative rotations.

from 0 up to various points, are shown marked off to illustrate that (for example) finding accidental agreement of two unrelated IrisCodes in 75% or more of their bits (a Hamming Distance of 0.25 or lower) has extremely small probability. As a consequence, we can tolerate a huge amount of corruption in iris images due to poor resolution, poor focus, occluding eyelashes and eyelids, contact lenses, specular reflections from the cornea or from eyeglasses, camera noise, etc. We can accept matches of very poor quality, say up to a third of the bits being wrong, and still make decisions about identity with the very high confidence levels indicated in Figure 6.

Decidability of Iris Recognition:

Finally, the overall "decidability" of the task of recognizing iris patterns is revealed by comparing the Hamming Distance distributions for "same" and for "different" irises. We have already seen the distribution obtained when different IrisCodes are compared, both before and after selecting the best match over multiple relative rotations. But when different images of the same iris are compared, acquired at different times and with various forms of corruption such as eyeglass reflections or scratches, camera noise, partial eyelid or eyelash occlusion, etc., we obtain the white distribution seen in Figure 7. (Note that identical images would produce Hamming Distances of 0.00, but the various sources of image corruption typically cause an identical iris in different images to have a Hamming Distance of 0.10 or higher.) Nevertheless, to the degree that one can confidently decide whether an observed sample belongs to the left or the right distribution in Figure 7, iris recognition can be successfully performed.

	Figure 7: Decision environment for recognition of iris patterns

For any such decision task the Decidability Index d' measures how well separated the two distributions are, since recognition errors are caused by any overlap between them. Even when two empirical distributions do not have any observed overlap, it is still possible to calculate their d' index since this is based upon the measured means and variances of the two distributions. The more they overlap, due either to having large variances or having their means close together (or possibly both), the lower the d' index will be. The definition of d' is: the difference between the means of the two distributions, divided by the square-root of their average variance. The higher d' is, the better: It indicates greater decision confidence. This measure of decidability (or detectability) is independent of how liberal or conservative the decision criterion is. Instead it reflects the degree to which any improvement in (say) the False Accept error rate must be paid for by a worsening of the False Reject error rate. The measured decidability is d' = 11.36 for iris patterns, which is much higher than for any other reported method of identifying persons.

By calculating the areas under the curves fitted to the observed pair of distributions of Hamming Distances, we can compute theoretical error rates as a function of the decision criterion employed. These are provided in the following Table, for various Hamming Distance acceptance thresholds:

		Figure 8: Theoretical probabilities of False Accept and False Reject errors, at various decision criteria for the acceptable Hamming Distances between enrolled and presenting IrisCodes. The cross-over point is 0.342, at which fraction of disagreeing bits the odds of either type of error are both equal to 1 in 1.2 million

Recognition Speeds:

Once an IrisCode has been computed, which takes about one second after image capture, it is compared exhaustively against all enrolled IrisCodes in the database, in search of a match. This search process is greatly facilitated by exploiting ergodicity (the representativeness of subsamples) and commensurability (the universal format and length of IrisCodes). On a Pentium processor the rate of raw comparisons approaches 100,000 IrisCodes per second, and this rate could be increased using dedicated PLA hardware to many millions of persons per second if such large databases of IrisCodes were ever enrolled.

Recognition versus Verification:

Because the probabilities of False Accepts are so low even at rather high Hamming Distances, as shown in the Table above, it is possible (and indeed routine) with this approach to perform exhaustive searches through very large databases for recognition of a presenting iris pattern, rather than merely a one-to-one comparison for verification. Clearly, exhaustive search recognitions are far more demanding than mere verifications, since the probabilities of a False Accept in any single comparison are increased proportionately with the size of the exhaustive search database. More precisely, if P_1 is the probability of a False Accept in a single (one-to-one) verification trial, then P_N, the probability of getting any False Accepts in recognition trials after searching exhaustively through a database of N impostors, is:

P_N = 1 - (1 - P_1)^(N)

This is a terribly demanding relationship. For example, even if P_1 were 0.001 (better than the existing performance of other biometrics), then even after searching through a database of merely N = 200 enrollees, the probability of getting one or more False Accepts among these impostors is P_N =0.181.

When the database of enrollees has grown merely to N = 2,000 the probability of a False Accept among them will have grown to P_N = 0.86. However, with iris-based recognition, the confidence levels against a False Accept are so high that we can afford to search even a nationwide (or even a planetary) database exhaustively, and still suffer only minuscule chances of a False Accept despite so many opportunities. The above table of cumulatives under the fitted British Telecom distributions indicates that if we use an acceptance Hamming Distance criterion of 0.28 (i.e. allowing up to 28% of the bits in two IrisCodes to disagree while still accepting them as a match), the False Accept probability in single trials is 10^(-12). Even after diluting down these odds by performing an exhaustive search over the total number of human irises on the planet, roughly 10^(10), the chances of a False Accept among them would still be only 1%. This is an extraordinary statistical situation for a recognition system, and it reveals the power of combinatorics to solve pattern recognition problems by reducing them to the detection of the failure of a test of statistical independence.

References:

Adler, F.H. (1965) Physiology of the Eye: Clinical Application. London: The C.V. Mosby Company.

Daugman, J. (1993) High confidence visual recognition of persons by a test of statistical independence. IEEE Trans. Pattern Analysis and Machine Intelligence, 15(11): 1148-1161.

Daugman, J. (1994) US Patent 5,291,560. Biometric Personal Identification System Based on Iris Analysis. Issue date: March 1, 1994.

Daugman, J., and Downing, J. (1995) Demodulation, predictive coding, and spatial vision. Journal of the Optical Society of America A, 12(4): 641 -660.

Mathematical Explanation of Iris Recognition

An "IrisCode" is constructed by demodulation of the iris pattern. This process uses complex-valued 2D Gabor wavelets to extract the structure of the iris as a sequence of phasors (vectors in the complex plane), whose phase angles are quantized to set the bits in the IrisCode.

This process is performed in a doubly-dimensionless polar coordinate system that is invariant to the size of the iris (and hence invariant to the imaging distance and the optical magnification factor), and also invariant to the dilation diameter of the pupil within the iris.

The demodulating wavelets are parameterized with four degrees-of-freedom: size, orientation, and two positional coordinates. They span several octaves in size, in order to extract iris structure at many different scales of analysis. Because the information extracted from the iris is inherently described in terms of phase, it is insensitive to contrast, camera gain, and illumination level (unlike correlation methods). The phase description is very compact, requiring only 256 bytes to represent each iris pattern. The 2D wavelets optimize the inherent Heisenberg-Weyl uncertainty relation for extraction of information in conjoint spatial - spectral representations.

The recognition of irises by their IrisCodes is based upon the failure of a test of statistical independence. Any given IrisCode is statistically guaranteed to pass a test of independence against any IrisCode computed from a different eye; but it will uniquely fail this same test against the eye from which it was computed. Thus the key to iris recognition is the failure of a test of statistical independence.

The equations and wavelet phasor diagram below summarize the pattern encoding process. The test of statistical independence on these complex phasors generates similarity metrics that are binomially-distributed. More detailed information about the complex-valued 2D Gabor encoding wavelets, about the test of statistical independence, and about the table of performance probabilities generated by IrisCodes, can be found in the published papers cited in the References at the bottom of this page.

Advantages of the Iris for Identification

Highly protected, internal organ of the eye

Externally visible; patterns imaged from a distance
Iris patterns possess a high degree of randomness
variability: 244 degrees-of-freedom
entropy: 3.2 bits per square-millimeter
uniqueness: set by combinatorial complexity
Changing pupil size confirms natural physiology
Pre-natal morphogenesis (7th month of gestation)
Limited genetic penetrance of iris patterns
Patterns apparently stable throughout life
Encoding and decision-making are tractable
image analysis and encoding time: 1 second
decidability index (d-prime): d' = 7.3 to 11.4
search speed: 100,000 IrisCodes per second
Disadvantages of the Iris for Identification
Small target (1 cm) to acquire from a distance (1 m)
Moving target ...within another... on yet another
Located behind a curved, wet, reflecting surface
Obscured by eyelashes, lenses, reflections
Partially occluded by eyelids, often drooping
Deforms non-elastically as pupil changes size
Illumination should not be visible or bright
Some negative (Orwellian) connotations

References

1. Daugman J (1999) "Wavelet demodulation codes, statistical independence, and pattern recognition." Institute of Mathematics and its Applications, Proc. 2nd IMA-IP. London: Albion, pp 244 - 260. (500K PostScript copy here.)
2. Daugman J (1999) "Biometric decision landscapes." Technical Report No. TR482, University of Cambridge Computer Laboratory. (285K PostScript copy here.)
3. Daugman J and Downing C J (1995) "Demodulation, predictive coding, and spatial vision." Journal of the Optical Society of America A, vol. 12, no. 4, pp 641 - 660.
4. Daugman J (1993) "High confidence visual recognition of persons by a test of statistical independence." IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 15, no. 11, pp 1148 - 1160.
5. Daugman J (1985) "Uncertainty relation for resolution in space, spatial frequency, and orientation optimized by two-dimensional visual cortical filters." Journal of the Optical Society of America A, vol. 2, no. 7, pp 1160 - 1169.

Anatomy and Physiology of the Iris

The iris is a protected internal organ of the eye, located behind the cornea and the aqueous humour, but in front of the lens. It is seen in cross-section in the anatomical drawing above. It is the only internal organ of the body that is normally visible externally. Images of the iris adequate for personal identification with very high confidence can be acquired from distances of up to about 3 feet (1 meter).

Among the visible features of an iris are the trabecular meshwork of connective tissue (pectinate ligament), the collagenous tissue of the stroma, ciliary processes, contraction furrows, crypts, rings, a corona and pupillary frill, colouration, and sometimes freckles. The striated anterior layer covering the trabecular meshwork creates the predominant texture seen with visible light, but all of these sources of radial and angular variation taken together constitute a distinctive "fingerprint" that can be imaged from some distance. Further properties of the iris that enhance its suitability for use in high confidence identification systems include: (i) its inherent isolation and protection from the external environment; (ii) the impossibility of surgically modifying it without unacceptable risk to vision; and (iii) its physiological response to light, which provides one of several natural tests against artifice.

A property the iris shares with fingerprints is the random morphogenesis of its minutiae. Because there is no genetic penetrance in the expression of this organ beyond its anatomical form, physiology, colour and general appearance, the iris texture itself is stochastic or possibly chaotic. Since its detailed morphogenesis depends on initial conditions in the embryonic mesoderm from which it develops, the phenotypic expression even of two irises with the same genetic genotype (as in identical twins, or the pair possessed by one individual) have uncorrelated minutiae. In these respects the uniqueness of every iris parallels the uniqueness of every fingerprint, common genotype or not. But the iris enjoys further practical advantages over fingerprints and other biometrics for purposes of automatic identification, including: (iv) the ease of registering its image at some distance from a Subject without physical contact, unintrusively and perhaps inconspicuously; (v) its intrinsic polar geometry, which imparts a natural coordinate system and an origin of coordinates; and (vi) the high level of randomness in its pattern, creating inter-Subject variability spanning 266 degrees-of-freedom, and an entropy (information density) of 3.4 bits per square-millimeter of iris tissue.

Developmental Morphogenesis and Chromatic Properties

The human iris begins to form during the third month of gestation. The structures creating its distinctive pattern are complete by the eighth month of gestation, but pigmentation continues into the first years after birth. The layers of the iris have both ectodermal and mesodermal embryological origin, consisting of (from back to front): a darkly pigmented epithelium; pupillary dilator and sphincter muscles; heavily vascularized stroma (connective tissue of interlacing ligaments containing melanocytes); and an anterior layer of chromataphores and melanocytes with a genetically determined density of melanin pigment granules. The combined effect is a visible pattern displaying various distinctive features such as arching ligaments, crypts, furrows, ridges, and a zigzag collarette. Iris colour is determined mainly by the density of the stroma and its melanin content, with blue irises resulting from an absence of pigment: long wavelength light penetrates and is absorbed by the pigment epithelium, while shorter wavelengths are reflected and scattered by the stroma. The heritability and ethnographic diversity of iris colour have long been studied, but until the present research, little attention had been paid to the achromatic pattern complexity and textural variability of the iris among individuals.

Probability of two irises producing exactly the same code: 1 in 10⁷⁸

Vein Biometrics

Blood vessel image maps have many of the advantages of the thermal image biometric. As indicated on the Vein Biometric Home page (http://innotts.co.uk/~joerice/), vein biometric systems can record subcutaneous infra red absorption patterns to produce unique and private identification templates for users. The technology is a vascular "bar" code reader for people. Veins and other subcutaneous features present large, robust, stable and largely hidden patterns. Subcutaneous features can be imaged within the wrist, palm, and dorsal surfaces of the hand. The technology can be applied to small personal biometric systems e.g. Biowatches and Biokeys and to generic biometric applications including intelligent door handles, door locks etc.

Vein pattern infrared grey-scale images are binarized, compressed and stored within a relational database of 2D vein images. Subjects can be verified against a reference template in under 200ms.

Vein Pattern Advantages

1. The human vascular structure is a unique & private feature of an individual.
2. Infra Red absorption patterns are easily compared via optical and DSP techniques.
3. Identical twins have different and distinct IR absorption patterns.
4. Uniqueness of vein patterns tested by Cambridge Consultants Ltd.
5. Veins provide large, robust, stable and hidden biometric features.
6. Vein patterns are not easily observed, damaged, obscured or changed.
7. Vein patterns require only low resolution IR. imaging allied to simple image processing.
8. Vein pattern stability and repeatability requires only simple algorithms for auto identification.
9. Vein structures provide the opportunity for low cost personal worn and pocket biometric keys.
10. Biometric Keys (Biokeys & Biowatches) read wrist or hands dorsal vein structures.
11. Biowatches output encrypted access codes whilst strapped to a recognized non coerced wrist.
12. Biokeys & Biowatches output encrypted access codes (Cryptographic signatures) to vehicles, computers, access portals, weapons, firearms etc.
13. Lost Vehicle Biokeys are not a problem; use any other! Vehicle Biokeys default to transmitting unrecognized patterns as plain text.
14. Vehicle security/engine management systems will let you in and drive if they recognize your vein pattern.
15. Biokeys & Biowatches maintain biometric privacy by placing the ownership of the biometric system, data & Crypto keys in the hands of the users.

Performance: Biometrics often fail, e.g. rain on face or glasses or fingerprints damaged by gardening or other DIY activity. Vein patterns are within you, are not left around (like fingerprints) nor can they be easily observed like iris patterns or faces. Also vein structures are not easily covertly captured or reproduced like other biometric traits.

Optical properties of tissue

Factors which affect the optical properties of tissue include pressure, temperature, anemia, blood loss, skin pigmentation and various diseases. Environmental factors can also play a role such as altitude and physical movement.

Data on tissue absorption coefficients, as a function of illuminating wavelength are not available with great precision and most sources of such information present it as graphs of the form already shared with ABI (see graphs I and II at the end of this report. As is evident from these graphs, in the visible and near infrared portion of the spectrum, it is scattering rather than absorption that dominates. Maximum penetration into the body occurs as follows:

These are optical penetration depths for soft tissue and represent 1/e depths. Clearly for maximum penetration with a readily available light source (e.g LED) red or infrared wavelengths are ideal. However, even using these wavelengths, it is important to realize that scattering limits penetration to only a few mm. This does not mean that there is no penetration at greater depths, it only means that the intensity is greatly diminished and scattering makes the interrogation of tissue at greater depths very challenging. Time gating with pulsed laser source or tomographic methods can be employed try to reveal deeper features, and this is of particular importance in clinical applications. If a more intense light source is employed, the diffuse scatter component becomes increasingly prominent.

Much work has been done to try to model light propagation in tissue and quite sophisticated models have been developed, such as those based on Monte Carlo techniques (see e.g. Farina et al, Physics in Medicine and Biology, Vol 44, Jan 1999, pp1 -11). These models incorporate scattering volume, scattering and absorption coefficients, anisotropy factors and refractive indices of the media involved. Tissue typically has an absorption coefficient of 0.5 cm-1, a scattering coefficient of 23 cm-1 and an anisotropy coefficient of 0.75 when illuminated at 633nm. It has been reported that in some diseased tissues (e.g. liver) lower anisotropy, scattering and absorption coefficients are observed, compared with healthy tissue (e.g. Germer et al, Lasers in Surgery and Medicine, Vol 23, 1998, pp 194-203), resulting in higher penetration depths. Moreover, certain tissues exhibit statistically significant increased penetration for relatively small increments in wavelength. For example, in some tissue 665nm has a 20% increase in penetration over 633nm.

From the literature, the wavelength used is one of the most relevant factors in the use of blood vessel maps for a biometric. It has been established that for oximeters, measuring the oxygen content of blood, that a two wavelength-based method provides quantitative accuracy. For accurate measurement of oxygen at low saturations, the use of 735nm and 890nm have been found to be optimal, for maximum penetration given tissue heterogeneity. At high saturations, 660nm and 900nm are preferred. (See for example Mannheimer et al, IEEE Trans on Biomedical Engineering, Vol. 44, 1997, pp 148-158).

Biometric Glossary