Li S.Z., Jain A.K. (eds.) Encyclopedia of Biometrics

Подождите немного. Документ загружается.

Latent Fingerprint

A ﬁngerprint left on an object by touching it. Example

objects are glasses, doors, and tables.

▶ Biometric Identification

▶ Fingerprint Features

▶ Latent Fingerprint Experts

▶ Law Enforcement Agency

▶ Liveness Detection: Fingerprint

▶ Security and Liveness, Overview

▶ Universal Latent Workstation

Latent Fingerprint Experts

THOMAS A. BUS EY,BETHANY L. SCHNEIDER

Psychological and Brain Sciences; Program in

Cognitive Science, Indiana University, Bloomington

Indiana, IN, USA

Synonym

Latent ﬁngerprint recognition

Definitions

Cognitive processing is the term given to mental effort

directed toward a particular problem. Cognitive science

is an umbrella term given to all disciplines that focus on

intelligent systems; research psychologists traditionally

focus on human performance. The ﬁeld of cognitive

science includes mathematicians, computer scientists,

research psychologists, biologists, and philosophers.

Cognitive processing is closely linked to perceptual pro-

cessing and decision making, both of which are involved

in latent print examinations. As part of the science,

researchers typically collect data from experts and

novices to document how and when expertise develops.

▶ Latent ﬁngerprint examiners are practitioners

who are trained to individualize or exclude latent

prints and prints from known sources (e.g., 10-print

cards). These practitioners often work with automated

databases such as IAFIS, which provide candidate

prints from known sources.

Introduction

Engineers attempt to solve a biometric problem by

isolating features or dimensions that they believe are

diagnostic, or use machine learning procedures to

identify a feature set that might be useful. Cognitive

scientists take the opposite approach. They use testing

procedures designed to infer the brain processes that

underlie performance in human experts. Under the

assumption that humans have the most ﬂexible infor-

mation processing system and can use different levels

of information, this reverse engineering approach

holds the promise of improved quantitative analyses

of ﬁng erprints. This entry summarizes the work that

has characterized performance in latent print exami-

ners, and describes how cognitive and vision scientists

design experiments to reveal the mechanisms under-

ling human latent print identiﬁcation.

The study of expertise in latent print examiners is a

relatively new ﬁeld, and only a few group has published

on the topic. However, there is a great deal of research

in related ﬁelds, and this research is described where it

applies to latent print examinations.

Empirical Evidence

Research on ▶ perceptual expertise in human experts

proceeds much in the same way that an engineer would

evaluate the performance of a biometric system .

Researchers generate candidate mechanisms that de-

scribe how an expert would accomplish a particular

task. These candidate hypotheses are based on an anal-

ysis of the information available, along with known

perceptual and memory constraints of humans. The

‘‘system’’ (in this case the human) is tested with a

recognition or memory task and performance measures

such as false match rate and false non-match rate can be

computed. Because the difﬁculty of a particular task

depends on the choice of materials, researchers often

compare performance from human experts against

those of human novices. Busey and Vanderkolk [1]

tested experts and novices in a ﬁngerprint fragment

890

Latent Fingerprint

matching task that was one of the ﬁrst major studies of

expertise in latent print examiners. Described below

are the details of this stud y, which not only illustrates

how experts differ from novices but provides an illus-

tration of how research question s are developed and

answered in Cognitive Science.

Stimuli such as those shown in Fig . 1 were pre-

sented brieﬂy to expert latent print examiners and

novices. A single print would be shown for 1/5th of a

second, followed by a pattern mask for either 200 ms or

5 s. Then two prints would be shown, one of which

exactly matched the studied print fragment. The test

prints could either be whole or partially masked to

simulate a latent print, and could be presented with

or without noise (which simulates the fact that some

latent prints are recovered on textured or marked

surfaces which adds visual noise).

The data in Panel B of Fig. 1 illustrates several

factors. First, experts perform better than novices, a

difference that is even more pronounced at longer

Latent Fingerprint Experts. Figure 1 Panel A: Stimuli used to test novices and latent print examiners [1]. Panel B:

Empirical data demonstrating improved performance overall for experts, better performance in noise, and evidence for

configural processing (see text for details).

Latent Fingerprint Experts

891

delays (not shown). This demonstrates that experts

may have better visual memories or the ability to

re-code visual information into verbal descriptions

which survive for longer intervals. Second, the struc-

ture of the design allows an investigation of

▶ conﬁg-

ural processing. The partially masked prints have

exactly half of the information in the full prints. Per-

formance from the partially masked prints can there-

fore be used to predict what performance should be in

the full print condition, using a model called probabil-

ity summation. The dashed lines show the prediction of

the probability summation model, and demonstrate

that experts exceed that prediction for the prints in

noise. This illustrates that experts are gathering more

information from the second half of the print once it is

added to the ﬁrst half than one would have expected

based on their performa nce on the ﬁrst half. Anothe r

way to view this is that for experts, the whole is greater

than the sum of the parts. One interpretation of these

results is that experts use the information from one

half to make better use of the information from the

second half when both are present.

Converging evidence for conﬁgural processing was

found using brain recording in a second experiment.

When visual stimu li are presented, neurons in the brain

ﬁre and give off electrical activity than can be recorded

by placing electrodes on the surface of the scalp. This

form of electroencephalography (EEG) allows research-

ers to moni tor the ongoing brain activity that is elicited

by a visual stimulus. While this technique has only

coarse spatial resolution due to the spreading nature

of electrical charges, it has excellent temporal resolu-

tion, on the order of millisecond accuracy.

Researchers using this particular technology have

noted that stimuli that are known to be processed

using conﬁgural or holistic mechanisms such as faces

and similar trained stimuli provide a signature of

this conﬁgural processing [2]. Voltage recorded from

the head and averaged over lots of trials provides

the event-related potential (ERP), and faces produce

a very distinctive feature over the left and right parie-

tal regions of the brain. This feature is a downward-

trending component that has an inﬂection at about

170 ms after stimulus onset. Figure 2 illustrates this

feature, which has been termed the N170. When the

stimulus is inverted, which has the effect of reducing or

eliminating conﬁgural processing, the N170 is delayed

and slightly more negative-going.

This signature of conﬁgural processing provides a

means to test whether exper ts process ﬁngerprints con-

ﬁgurally. Experts and novices were shown upright and

inverted faces and ﬁngerprints. As expected, both

experts and novices show differences between upright

and inverted faces. However, only experts showed a simi-

lar pattern for ﬁngerprints as they showed for faces: The

N170 for inverted ﬁngerprints was signiﬁcantly delayed

relative to upright ﬁngerprints, but only in experts.

The dark curves in Fig. 2 illustrate this effect. These

Latent Fingerprint Experts. Figure 2 Electrophysiological data from Novices and Experts with upright and inverted

fingerprints and faces [1]. Light curves come from faces, while dark curves are from ﬁngerprints (see text for details).

892

Latent Fingerprint Experts

ﬁndings are important because not only do they illus-

trate demonstrably different patterns of brain activity

in experts and novices, but the time course of the

differences are consistent with processing that happens

relatively early in visual processing. Thus these exam-

iners are experiencing relatively low-level changes in

their visual system that improve the quality of the

information and the way they interpret this perceptual

information.

In summary, the behavioral and electrophysiologi-

cal evidence from latent print examiners supports the

view that experts have better recognition overall for

ﬁngerprints, they have better visual memories for ﬁn-

gerprint informati on, and they process ﬁngerprint

information in qualitatively different ways using con-

ﬁgural processing mechanisms.

Perceptual Expertise

While little research has focused speciﬁcally on latent

print examiners and the changes that develop as a result

of their expertise, candidate mechanisms that have been

previously discovered by cognitive scientists using

related materials can be extended. For example, the

idea of conﬁgural (using relational information between

parts) and/or holistic (obligatory processing of all the

parts of an object) processing has become a consistent

theme throughout the literature and many researchers

argue that it is a signature of expertise [3]. Speciﬁcally,

researchers studying perceptual expertise have devel-

oped paradigms that test for and illustrate a shift from

a feature-based system of object recognition (seeing

individual parts of an objects) to the use of holistic

and/or conﬁgural mechanisms [3]. These effects are

often illustrated in behavioral tasks that train subjects

on a speciﬁc stimulus type and then test these subjects

on either the studied or transformed conﬁguration or

isolated parts. Post-training performance is often com-

pared with either their pre-training performance or

with novices (those who receive no training). The un-

derlying theme that results from these research para-

digms is that experts develop a holistic system which

causes them to be more sensitive to conﬁgurations and

be unable to ignore distractor parts of the stimulu s.

Apart from establishing conﬁgural and/or holistic

mechanisms, another key issue in exp ertise studies is

showing how experience with a domain causes a reor-

ganization of the visual recognition hierarchy away from

the basic level and to the subordinate level. In general

terms, subjects more readily identify items based on the

basic level category membership (e.g., bird, table) rather

than their subordinate membership (e.g., robin, coffee

table) [4]. This hierarchy is structured to reﬂect the

prominent use of basic-level information over the sub-

ordinate level information. However, a series of experi-

ments has shown that the development of expertise

results in enhanced subordinate level identiﬁcation

[5]. It has been proposed that (1) expertise causes a

shift in the hierarchy to the subordinate level rather

than the basic level, (2) experts make identiﬁcations

based on this subordinate level information, and

(3) their expertise allows them to be equally proﬁcient

in making identiﬁcation on the subordinate and basic

level. The proﬁciency in which experts use this subor-

dinate level information has been reliably replicated

and has been argued to be a signature of expertise.

Other studies have researched differences between

experts and novices in terms of how expertise impac ts

▶ visual memory, the ability to use verbal redescrip-

tions, and attention to particular features. For exam-

ple, previous stud ies on expertise have implicated

enhanced visual memory for expertise items, and

showed that chess masters were able to accurately

reproduce VALID board conﬁgurations after viewing

them for only 5 s [6]. This is arguably due to their

extensive knowledge of speciﬁ c patterns that results

with expertise in the domain. Such an idea can be

applied to latent print examiners and has also been

reported for experts in other domains such as bridge

players [7], music students [8], and electronics techni-

cians [9]. In addition, this idea can be extended into

the category learning literature by a ﬁnding that shows

increases in memory sensitivity account for the ability

to learn to uniquely identify similar objects [10].

This enhanced memory ability could also be linked to

an enhanced ability to ﬁxate on features that are the

most informative for future identiﬁcation, an idea that

is also supported in the category learning literature [11].

In addition to visual memory, research in the per-

ceptual categorization literature argue that experts de-

velop a more robust storage, such as implicit verbal

redescriptions, in the process of specializing in a cate-

gory. Speciﬁcally, experts appear to garner more verbal

knowledge about a domain but make categorizations

without explicit deliberations [12].

Research with radiologists suggests that expertise

may alter what types of perceptual information are

Latent Fingerprint Experts

893

allowed for consideration. Speciﬁcally, experts argu-

ably attenuate to speciﬁc task relevant dimensions

[13]. More generally, this idea has also been founded

in the category learning literature by showing that

category learning includes learning how to optimally

allocate attention to those features relative to the cate-

gory and/or task and discard unrelated features [11].

Decision Making and Decision Biases

Expertise brings special abilities, but it also can lead

to special vulnerabilities. Several studies conducted

by Itiel Dror and co-workers illustrate the role that

context plays in decisions about ﬁngerprint individua-

lizations. This can bring about

▶ contextual biases.

The difﬁculty with latent print examinations is that

the judgment that two prints come from the same

source is essentially based on similarity. Even if one

source (say a ﬁngerprint) maintains a persistent struc-

ture over time, the way that print is laid down can

greatly affect its appearance. Thus no two impressions

of a single source will look identical. Clear prints

might look very similar, but in the end individualization

essentially comes down to a judgment that two prints

look more similar than any close non-match that

the examiner has seen. Such a task has three possible

decisions (individualization, exclusion, or insufﬁcient

detail to make a determination). A particular pair of

prints will produce some amount of evidence for each

of these decisions, but whether the evidence exceeds

some internal threshold depends on the individual

examiner. Dror et al. explored the possibility that

the details of the case (the context) might affect the

decision process.

In the ﬁrst study [14], non-experts where shown

pairs of ﬁngerprints and given additional (ﬁctitious)

details about the case. When the prints were shown

with a highly emotional context such as an accident

scene picture or a murder victim, the stimulus affected

the decision made by subjects. Subjects were more

likely to report a matching ﬁngerprint pair when the

context was emotional. This suggests that contextual

information beyond the particular ﬁngerprint percep-

tual information plays a role in latent print examina-

tions, at least with novices.

To extend this to experts, two additional studies

used covert measures to assess the role of context with

examiners during their normal workﬂow [15, 16 ].

To highlight the importance of this work, consider

the task of an examiner. He or she must evaluate the

perceptual evidence and decide whether there is sufﬁ-

cient evidence to make a decision. What constitutes

‘‘sufﬁcient’’ is of course of primary importance. Re-

search psychologists refer to this task as a criterion-

based (as opposed to a criterion free) judgment, since

the decision outcome is based in part on the criterion

that the examiner establishes. If the examiner allows

additional details about the case that are irrelevant to

the particular identiﬁcation at hand to inﬂuence their

criterion, they reduce value and independence of the

latent print examination.

In this particular set of studies, latent prints

from closed casework were given again to the same

experts under the guise of a new case. These prints

had previously been matched or excluded by the

examiners. Dror and his colleagues found that 8 out

of 11 experts made a decision that was inconsistent with

their previous decisions on the identical pairs of prints.

Most of the switched decisions occurred with difﬁcult

prints that were previously judged as identiﬁcations,

although some of these easy identiﬁcations also had

changed answers. The details of these experiments

are complex, and the reader is referred to the ori-

ginal sources for full details [17], but the implications

are clear: context can play a role in the decision that

an examiner makes, and care must be taken not

to allow external inﬂuence to affect the perceptual

judgment.

Summary

Research on latent print examiners has demonstrated

increased recognition of latent prints and the ﬂexibility

to rely on different levels of print information. These

differences are supported by superior visual memory

and different styles of process. The EEG data suggest

that training and experience are causing changes in

relatively early and low-level areas of the visual system

that improve the quality of the perceptual information.

A host of related studies suggest that experts may learn

to re-code visual information into verbal descriptions,

and learn to attend to the most relevant and diagnostic

regions of a print. However, with these increased abil-

ities may come increased vulnerabilities such as con-

textual biases that may affect the interpretation of a

ﬁngerprint pair.

894

Latent Fingerprint Experts

Related Entries

▶ Fingerprint Classiﬁcation

▶ Fingerprint Matching, Manual

▶ Fingerprint Recognition, Overv iew

▶ Law Enforcement

▶ Multiple Experts

▶ Psychology of Gait and Action Recognition

References

1. Busey, T.A., Vanderkolk, J.R.: Behavioral and electrophysiologi-

cal evidence for conﬁgural processing in ﬁngerprint experts. Vis.

Res. 45, 431–448 (2005)

2. Rossion, B., Gauthier, I.: How Does the Brain Process

Upright and Inverted Faces? Behav. Cogn. Neurosci. Rev. 1,

63–75 (2002)

3. Bukach, C.M., Gauthier, I., Tarr, M.J.: Beyond faces and mod-

ularity: The power of an expertise framework. Trends Cogn. Sci.

10, 159–166 (2006)

4. Rosch, E., Mervis, C.B., Gray, W.D., Johnson, D.M., Boyes-

Braem, P.: Basic objects in natural categories. Cogn. Psychol.

8, 382–439 (1976)

5. Tanaka, J.W., Taylor, M.: Object categories and expertise: Is the

basic level in the eye of the beholder? Cogn. Psychol. 23, 457–482

(1991)

6. Chase, W.G., Simon, H.A.: Perception in chess. Cogn. Psychol.

4, 55–81 (1973)

7. Charness, N.: Components of skill in bridge. Canad J Psychol.

33, 1–16 (1979)

8. Beal, A.L.: The skill of recognizing musical structures. Memory

Cogn. 13, 405–412 (1985)

9. Egan, D.E., Schwartz, B.J.: Chunking in recall of symbolic draw-

ings. Mem. Cogn. 7, 149–158 (1979)

10. Nosofsky, R.M.: Attention and learning processes in the identi-

ﬁcation and categorization of integral stimuli. J Exp Psychol

Learn, Mem. Cogn. 13, 87–108 (1987)

11. Zhang, L., Cottrell, G.W.: A computational model which learns

to selectively attend in category learning. Proceedings of the

2005 International Conference on Development and Learning,

vol. 19. pp. 195–200

12. Johansen, M.K., Palmeri, T.J.: Are there representational

shifts during category learning? Cogn. Psychol. 45, 482–553

(2002)

13. Sowden, P.T., Davies, I.R.L., Rolings, P.: Perceptual learning of

the detection of features in X-ray images: A functional role for

improvements in adults’ visual sensitivity? J. Exp. Psychol.:

Human Percep. Perform. 26, 379–390 (2000)

14. Dror, I.E., Peron, A.E., Hind, S.L., Charlton, D.: When emotions

get the better f us: The effect of contextual top-down processing

on matching ﬁngerprints. Appl. Cogn. Psychol. 19, 799–809

(2005)

15. Dror, I.E., Charlton, D.: Why experts make errors. J. Forensic

Identif. 56, 600–616 (2006)

16. Dror, I.E., Charlton, D., Peron, A.E.: Contextual information

renders experts vulnerable to making erroneous identiﬁcations.

Forensic Sci. Int. 156, 74–78 (2006)

17. Dror, I., Rosenthal, R.: Meta-analytically Quantifying the Reli-

ability and Biasability of Forensic Experts. J. Forensic Sci. 53,

900–903 (2008)

Latex Finger

▶ Fingerprint Fake Detection

Law Enforcement

KEN MOSES

Forensic Indentiﬁcation Services, San Franscisco, USA

Synonyms

Criminal law enforcement; Law enforcement agency;

Police law enforcement

Definition

Law enforcement systems record information about

individuals, both biometric data and demographic

data, that is used for quickly identifying criminals

and their acts.

Introduction

The identiﬁcation of criminals and repeat offenders

has been one of the primary missions of police agencies

since their creation in the nineteenth century. The

industrial revolution led to mass migration to cities

where anonymity gave safe harbor to lawbreakers.

Newspapers fanned both real and perceived threats to

public safety and placed persistant demands upon the

police to solve crimes. Laws were enacted that dealt the

Law Enforcement

895

harshest penalties to repeat offenders encouraging law-

breakers to adopt a different name (known as an alias)

with each new arrest.

Answering the call for effective systems of identiﬁ-

cation, law enforcement agencies initiated various sys-

tems to maintain records at the end of the nineteenth

century. One of the earliest techniques utilized the new

science of photography to take pictures or mug shots at

the time of arrest. These mugs shots were organized

into albums, called Rogues Galleries. The method of

recording these photographs was adopted and standar-

dized worldwide to include a front and proﬁle view of

the subject. General physical descriptions, including

scars, marks, and tattoos accompanied the photographs.

Today, Rogues Galleries have evolved into large digitized

databases of mug shots from which investigators can

select possible suspects based on victims’ descriptions,

or compile ‘‘six packs’’ of persons sharing similar

descriptions for use in photo line-ups [1]. Automated

facial recognition would be a natural extension of the

mug shot systems, and research is underway to build

that capability into the system that would be most

valuable in video surveillance cases.

In 1880, Alphonse Bertillon, a clerk in the Paris

police, devised an anthropomorphic system of identi-

ﬁcation based upon 18 body measurements utilizing

specialized rulers and calipers for precision. From these

measurements, Bertillon derived a numeric value or

‘‘portrait parle,’’ that was supposedly unique to each

person. Bertillon’s Portrait Parle met with great success

in identifying repeat offenders regardless of what

names they had assumed at the time of arrest. Within

a decade, this manual biometric system was in use in

police agencies worldwide (Fig. 1).

Law Enforcement. Figure 1 Early police biometrics circa. 1906 Bertillon ‘‘Portrait Parle’’ record.

896

Law Enforcement

During this period, a new type of biometric

was being explored by William Herschel in India and

Dr. Henry Faulds in Japan. In 1892, Francis Galton,

the nephew of Charles Darwin, published Finger

Prints [2], a book that became the standard reference

on the subject. Sir Edward Henry, police commissioner

in Bengal India, devised a classiﬁcation system that

allowed for easy search and retrieval of ﬁngerprints.

The Henry System was later adapted by Scotland

Yard in 1901. For several years, police agencies around

the world utilized both the Bertillon Portrait Parle

and ﬁngerprinting systems side by side until ﬁnger-

prints won out and Portrait Parle was largely aban-

doned. Fingerprints proved easier to record, ﬁl e, and

retrieve, and they could be used to solve crimes when

they were inadvertently left by criminals at the scene of

crime [3].

Today, as with mug shots, ﬁngerprints are usually

digitally recorded, stored, and automatically searched

by Automated Fingerprint Identiﬁcation Systems

(AFIS) [4]. AFIS has become the backbone of criminal

identiﬁcation and is the most widely used biometric

system in the world. Standards have been adopted

for the digital storage and transmission of ﬁngerprint

images [5].

In Finger Prints, Galton highlighted a number of

characteristics of each ﬁngerprint that are used to

uniquely identify it. Modern automated systems use

some of these Galton characteristics in extracting in-

formation from the print (Fig. 2).

Fingerprint Functions in Law

Enforcement

Law enforcement AFIS systems are composed of two

interdependent subsystems: the tenprint or criminal

ID subsyste m, and the latent or criminal investigation

subsystem. Each subsystem operates with a consider-

able amount of autonomy, and both are vital to public

safety.

The Tenprint subsystem is tasked with identifying

sets of inked or scann ed ﬁngerprints incident to an

arrest or citation or as part of the applicant process to

determine whether or not the person has an existing

record. An automated tenprint search usually involves

a search of only two or four ﬁngers. Submitted ﬁnger-

prints by and large have sufﬁcient clarity and detail to

make searching of more than two ﬁngers redundant. At

present the AFIS can return a search of one million

records in under a minute.

The latent print or criminal investigation subsys-

tem is tasked with solving crimes through the identiﬁ-

cation of latent prints developed from the scene of

crime. Terminals used within the latent subsystem

are often specialized to accommodate the capture

and digital enhancement of individual latent prints

(Fig. 3).

The search of a latent print is more tedious and

time consuming than a tenprint search. Latent prints

are often fragmentary and of poor image quality.

Minutiae features must be reviewed by a latent print

Law Enforcement. Figure 2 Computer plotting of minutiae or Galton details from a fingerprint.

Law Enforcement

897

examiner one by one before the search even begins.

Once the search is launch ed, the AFIS searches the

database and returns a respondent list of the closest

matches [6–8]. The ﬁngerprint examiner then com-

pares each candidate against the search print.

The id entiﬁcation of latent ﬁngerprints from the

scenes of crime has been a tremendous boon to law

enforcement agencies. AFIS identiﬁcations, or hits, are

responsible for solving hundreds of thousands

of crimes each year throughout the world. In 2005,

in the United States alone, 50,000 cases were cleared

by AFIS hits.

References

1. The Criminal Investigative Process; Rand, R-1776-DOJ; October

(1975)

2. Galton, F.: Finger Prints. Macmillan, London (1892)

3. Cole, S.: Suspect Identities. Harvard University Press,

Cambridge, MA (2001)

4. Lee, H.C., Gaensslen, R.E.: Advances in Fingerprint Technology,

2nd edn. CRC Press, West Palm Beach, FL (2001)

5. Data Format for the Interchange of Fingerprint, Facial, & Scar

Mark & Tattoo (SMT) Information (ANSI/NIST-ITL 1-2000)

6. Komarinski, P.: Automated Fingerprint Identiﬁcation Systems.

Elsevier, Amsterdam (2005)

7. Ratha, N., Bolle, R.: Automated Fingerprint Recognition Sys-

tems. Springer, New York (2004)

8. Maltoni, D., et al.: Handbook of Fingerprint Recognition.

Springer, New York (2003)

Law Enforcement Agency

▶ Law Enforcement

LBP (Local Binary Pattern)

▶ Local Image Filters

LCN DNA/Low Template Level

Low template level or as formally referred to as Low

Copy Number (LCN) DNA refers to the analysis of

trace amounts of DNA (typically less than 100 pg). As

LCN is an ultrasensitive technique, there are artefacts

that one must be aware (allele drop-in and drop-out).

It requires careful interpretation and very strict proto-

cols to avoid contamination.

▶ Forensic DNA Evidence

Law Enforcement. Figure 3 Latent fingerprint workstation.

898

Law Enforcement Agency

LDA (Linear Discriminant Analysis)

Linear discriminant analysis is deﬁned as an orthogo-

nal linear transformation that best separate two or

more classes of objects. It ﬁnds the set of the projection

vectors which maximize the ratio of between-class

scatter against within-class scatter. The resulting com-

bination may be used for classiﬁcation.

▶ Face Variation

▶ Illumination Compensation

▶ Linear Dimension Reduction

LED (Light Emission Diode)

LED refers to an electronic component, Light Emission

Diode, which emits light when traversed by a current

of certain entity.

A light-emitting diode, also called an LED, is a

semiconductor diode that converts electric energy into

electromagnetic radiation at a visible and near infrared

frequencies when its p–n junction is forward biased.

▶ Biometric Sensor and Device, Overview

▶ Face Recognition, Near-infrared

Lighting Compensation

▶ Illumination Compensation

Lighting Model

The realism of a computer generated visual scene

depends on the extent to which lighting phenomena

are mimicked by the lighting model used. In Computer

Graphics, the lig hting model refers to the choice of

reﬂectance function used. However, in Computer Vi-

sion, it may also include the t ype of light source(s) and

of normals used, as these have a considerable effect on

the synthesis and on the analysis by synthesis results. A

reﬂectance function is a mathematical function that

describes how light is reﬂected from an object.

▶ Face Sample Synthesis

Likelihood Ratio Test

The likelihood ratio test results from simplifying the

ratio between the probability that the sensor’s observa-

tion resulted from the positive hypothesis (genuine

user) divided by the probability that the sensor’s ob-

servation resulted from the negative hypo thesis (im-

poster). The objective is to improve the ratio by

making the probability distributions separate as

much as possible. If the distributions are far apart,

the probability that the observation came from the

genuine user increases, and the probability that the

observation came from the imposter decreases. This

increases the ratio improving the test.

▶ Fusion, Decision Level

Limbus

The limbus is the outer boundary of the annular iris

structure, where the iris meets the white portion of the

eye, the sclera.

▶ Iris Image Data Interchange Formats, Standardization

Linear Dimension Reduction

WEI-SHI ZHENG

,J.H.LAI

,PONG C. YUEN

Sun Yat-sen University, Guangzhou, P.R. China

Hong Kong Baptist University, Hong Kong, China

Synonym

Linear feature extraction

Linear Dimension Reduction

899