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

Подождите немного. Документ загружается.

the users or for statistical purposes, are a combination

of two or three minutiae of basic type.

The minutiae contribute the most to the selectivity

of the ﬁngerprint, due to the combination of their

spatial arrangeme nt along the ridges and their intrinsic

characteristics: type, location, and orientation. The

selectivity offered by a minutiae conﬁguration present

on a ﬁngerprint or on a ﬁngermark is a function of

their number, ty pe, and topology (relative position and

orientation on the ridges).

The process underlying the development of the

minutiae is not known yet, but models offered by

mathematical biology and empirical studies suggest

that it is epigenetic [2]. For ridge endings, bifurcations,

and dots, more correlations are observed on ﬁnger-

prints of monozygotic twins as opposed to dizygotic

twins [4]. Correlations are also observed between the

number of minutiae and the ﬁnger number, which can

be explained by the fact that the surface of the ﬁngertip

of the thumb is bigger than the surface of the ﬁnger-

tip of the little ﬁnger. The relative frequencies of

the minutiae type are correlated with gender, but no

difference has been observed between the ﬁngerprint

characteristics of the left and right hands [5].

Third Level Details

The study of the friction ridge details may be further

subdivided into the description of ridge contours or

edges, and the position and the shape of the pores [6].

However, the degree of agreement between dactylosco-

pists on the value of these latter characteristics is

limited so far, and no systematic study supports the

different opinions.

Statistical Research

The ﬁrst statistical investigations were conducted at the

end of the nin eteenth and at the beginning of the

twentieth century, but the initial models were devel-

oped on the basis of unrealistic premises: it was pre-

sumed that each minutiae type appeared with the same

probability and independently of each other on the

ridge skin surface. More sophisticated models were

developed later during the twentieth century, ﬁrst in-

cluding the unbalance between the minutiae type (e.g.,

the bifurcations are more rare than the ridge endings)

and then including the uneven density of the minutiae

(e.g., the density of minutiae increases in the centre

and delta zones) [7].

Statistical studies mainly focus on the second level

features and especially the spatial arrangement of minu-

tiae, while studies of other ﬁngerprint features remain

too seldom. These studies on minutiae provide extreme-

ly valuable fundamental knowledge about the degree

of randomness of minutiae conﬁgurations, but they

cannot be used yet for the deployment of large-scale,

case-speciﬁc statistical evaluation of the ﬁngermark

evidence. Current statistical models simplify reality,

emphasizing the statistical behavior of minutiae, and

adopting a restricted view of the overall factors like the

general pattern, the main ridge ﬂows, the ridge edges,

or the pores. Nevertheless, this new approach aims to

offer a uniform framework and a transparent method-

ology to the dactyloscopists. Coupled to a logical infer-

ence model originating in the Bayes theorem, these

models aim to assist them in producing a logical, test-

able, and quantitative evaluation of the ﬁngerprint

evidence based on statistical probabilities [8].

Fingerprint, Forensic Evidence of. Figure 2 The three basic structures of minutiae: the ridge ending, the bifurcation,

and the dot.

530

Fingerprint, Forensic Evidence of

Classification of Fingerprints and

Fingermarks

Manual Classification

For about a century the classiﬁcation of ﬁngerprints

based on general patterns allowed the dactyloscopists

to limit the search for the source of an unidentiﬁed

ﬁngermark to a speciﬁc section of their databases of

ﬁngerprint reference ﬁles. Francis Galton proposed the

ﬁrst system of ﬁngerprint classiﬁcation in 1891, and

the development and practical application of dactylo-

scopy for forensic use were materialized in 1892 with

the publication of his manual of dactyloscopy. This led

to the acceptance of ﬁngerprints in Great-Britain and

the British Empire. In 1900, Henry modiﬁed the clas-

siﬁcation system of Galton, which remained the most

widely used system in the world under the name of

Galton-Henry. In 1891, Vucetich began to collect the

ﬁrst ten print cards databases based on the ideas of

Francis Galton and developed another classiﬁcation

system, which was adopted by some South-American

countries. The size of the ten print cards databases

increased progressively during the twentieth century,

and the workability was maintained sophisticating the

indexation system, but to the cost of a trade-off be-

tween selectivity and reliabilit y. The coexistence of

several classiﬁcation systems around the world limited

the interoperability of the manual classiﬁcation be-

tween different systems. In the second part of the

twentieth century, manual classiﬁcation was slowly

abandoned and replaced by computerized classiﬁ-

cation systems named Automatic Fingerprint Identi-

ﬁcation Systems (AFIS) [9].

Automatic Classification

Development

From the mid-1960s, research on automation of

ﬁngerprint identi ﬁcation started. USA and Japan con-

centrated on automation of the high-volume ten-print

workload, while France and the UK focused more on

automation of ﬁngermark identiﬁcation. After a de-

cade of effort, digitization of the ten-print card and

automatic designation of minutiae were effective

enough for the USA and the UK to produce auto-

matic ﬁngerprint reader systems. This advancement

opened the possibility to digitize the ten print card

records and to store the standard impressions and

the demographic data of individuals (e.g., name,

citizenship, and date of birth) in a computerized

database.

Forensic Uses of AFIS Technology

AFIS technology was initially developed to assist the

dactyloscopists with computers in the identity veriﬁca-

tion process of individuals through their ﬁngerprints.

This process consists in searching the ten ﬁngerprints of

an individual in the database of standard impressions to

verify if he or she is already present in the database and,

if present, to check his or her demographic data. The

AFIS technology has achieved enough maturity to en-

sure an identity veriﬁcation process that is virtually

error-free from the technological point of view, even if

clerical mistakes in the database or in the running of the

process can never be excluded.

In the 1990s the improvement of both AFIS and

computer technologies allowed for the processing of

ﬁngermarks, exploited in two forensic processes. Fin-

germarks can be used for forensic investigation, in

order to establish a link between a crime scene or an

object and an individual. They can also be used for

forensic intelligence to establish links between several

crimes, even if the potential for links using marks

depends on their limited quality.

In the 2000s the improvement of the computer

mass-storage, in terms size and affordabilit y, favored

the constitution of large-scale palmprints databases.

This development allowed for an extension of forensic

investigation and forensic intelligence based on palm-

marks. In most countries, the constitution of large

scale palmprints databases is an ongoing process.

The challenge of standardization has only been

solved recently, through the use of a common format,

developed by the American National Institute for Stan-

dards and Technology (NIST), facilitating the compu-

terized exchange of ﬁngerprint and ﬁngermark data

between countries and agencies [10].

Individualization of Fingerprints and

Fingermarks

History

The criminalist Edmond Locard enounced the

ﬁrst rule establishing a minimum number of minutiae

Fingerprint, Forensic Evidence of

531

necessary for ﬁngermark identiﬁcation. During

1911–1912 he initiated the discussion of a numerical

standard for the forensic identiﬁcation of ﬁngermarks,

suggesting the following rule:

1. If more than 12 minutiae (‘‘concurring points’’) are

present, and the ﬁngermark is sharp, then the iden-

tity is certain. The imperative requirement for the

absence of signiﬁcant differences is imp licit.

2. With 8–12 concurring points, the case is borderline

and the certainty of the identity depends on

a. The sharpness of the ﬁngermark.

b. The rarity of the type.

c. The presence of the core of the general pattern

and the delta in the usable part of the mark.

d. The presence of pores.

e. The perfect and obvious similarity of the print

and the mark regarding the width of the papil-

lary ridges and valleys, the direction of the lines,

and the angular value of the bifurcations.

In these instances, the certainty of the identiﬁca-

tion can only be established following a discussion

of the case by at least two competent and experi-

enced specialists.

3. With less than 8 minutiae, the ﬁngermark cannot

provide certainty for the identiﬁcation, but only a

presumption proportional to the number of min-

utiae available and their clarity.

Principally the ﬁrst two parts of this rule were

largely adopted by the community of the dactylosco-

pists but, unfor tunately, the third part of the rule

remained largely ignored [5].

Current Practice

The current dactyloscopic practice has evolved from

the body of knowledge developed about the ﬁngerprint

individuality and the forensic use of ﬁngermarks.

It is formalized in a 4-step procedure named ACEV

(Analysis-Comparison-Evaluation-Veriﬁcation).

This procedure consists in the analysis of the ﬁngermark

followed by the analysis of the ﬁngerprint, the compari-

son of the ﬁngermark and the ﬁngerprint, the evalua-

tion and the decision based on the observed similarities

and discrepancies between the ﬁngermark and the ﬁn-

gerprint, and the veriﬁcation of the ﬁndings by a second

dactyloscopist.

Despite the formalization of the identiﬁcation pro-

cedure, the practice varies between continents and

countries, and even within some countries. The evalu-

ation step, in particular, is based either on a quantita-

tive threshold or on a qualitative threshold.

Quantitative Threshold: Presence of a Numerical

Standard

A majority of European and South American

cou ntries favor a purely quantitative a pproach for

forensic indiv iduali zation, by ﬁxing a numerical

standard and considering qualitative aspects such

as the third level details as secondary. A formal iden-

tiﬁcation is established only if a minimal number

of corresponding m inutiae between the obser ved

mark and the ﬁngerprint – and an abse nce of signiﬁ-

cant diffe rences – is put in evidence.

The numerical standard differs between countries

and sometimes also between agencies in the same coun-

try: Italy (16-17); UK (before 2000) (16); Belgium,

France, Israel, Greece, Poland, Portugal, Romania,

Slovenia, Spain, Turkey, South American Countries

(12); Netherlands (10-12); Germany (8-12); Switzerland

(before 2008) (8-12); and Russia (7) [5].

Qualitative Threshold: Absence of Numerical Standard

Until 1970, the ﬁngerprint identiﬁcation procedure in

the USA was also based on a numerical standard of

12 points, and below this threshold, qualitative factors

in the comparison were taken into consideration. In

1970, a commission of experts from the Inte rnational

Association for Identiﬁcation (IAI) was established to

study the question of the relevancy of a ﬁxed numerical

standard for dactyloscopy. The following resolution

was adopted by the IAI in 1973: ‘‘The International

Association for Identiﬁcation, based upon a 3-year

study by its Standardization Committee, hereby states

that no valid basis exists for requiring a predetermined

minimum of friction ridge characteristics that must be

present in two impressions in order to establish posi-

tive identiﬁcation.’’

It was accepted that the concept of identiﬁcation

could not be reduced to counting ﬁngerprint minutiae,

because each identiﬁcation process represents a unique

set of features available for comparison purposes; the

identiﬁcation value of concurring points between a

ﬁngerprint and a ﬁngermark depends on a variety of

conditions that automatically excludes any mini mum

standard.

532

Fingerprint, Forensic Evidence of

In 1995, during a conference meeting on ﬁnger-

mark detection techniques and identiﬁcation hosted in

Ne’urim, Israel, 28 scientists active in the ﬁeld of dac-

tyloscopy, representing 11 countries, unanimously ap-

proved a resolution that is a slight variation of the IAI

1973 resolution. The Ne’urim declaration states that

‘‘no scientiﬁc basis exists for requiring that a predeter-

mined minimum number of friction ridge features

must be present in two impressions in order to estab-

lish a positive identiﬁcation.’’

Decision Process

A formal identiﬁcation is established when the dacty-

loscopists reach a decision threshold. They evaluate the

contributions to individuality on a quantitative level

(numerical standard), or on a qualitative level (absence

of numerical standard), and the size of the relevant

population of potential sources of the ﬁngermark is set

to its maximum, independently of the circumstances of

the case [5].

On the basis of their evaluation, most dactylosco-

pists report three types of qualitative opinion: identiﬁ-

cation, exclusion, and inconclusive. As their evaluation

is deterministic, they also make an implicit use of

their own subjective probabilities of the rarity of the

characteristics used to substantiate their opinion. They

reﬁne these subjective probabilities through training

and experience, but they rarely consider results from

research, particularly in the ﬁelds of embryology and

statistics.

Admissibility of the Fingerprint in the USA

Like for other forensic disciplines, the scientiﬁc status

of ﬁngerprint identiﬁcation has been questioned since

1993, when the Supreme Court of the USA handed

down its ruling in Daubert v. Merrell Dow Pharma-

ceuticals (1993, Inc., 509 US, 579). Previously the main

criterion for the admissibility of expert testimony in

the federal courts of the USA was the Frye standard,

which requires the general acceptance of the methods

by the relevant scientiﬁc community. Daubert gave

federal judges much greater discretion in deciding ad-

missibility. It suggested that they consider (1) whether

a theory or technique can be tested, (2) whether it has

been subject to peer review, (3) whether standards exist

for applying the technique, and (4) the technique’s

error rate. Although it is possible to test and validate

methods for the forensic individualization of ﬁnger-

marks, the research on this topic is still very limited.

The admissibility of ﬁngerprint evidence, as being

scientiﬁc in nature, has been subject to a Daubert hearing

in the case U .S. v. Mitchell (1999, U.S. District Court for

the Easte rn District of Pennsylvania, Criminal), followe d

by Daubert hearings in mor e than 20 other ﬁngermark

cases. In the same case, U.S. v. Mitchell, the FBI provided

calculations based on experiments carried out on an AFIS

system. Random match probabilities of 10

97

and 10

27

were claimed respectively for complete ﬁngerprints

and partial ﬁngermarks. These extraordinary numbers

have been obtained by an extreme extrapolation of the

probability density of the score using a postulated

model, but they are so far from reality that it is

surprising that they were admitted as evidence. Until

January 2002, all Daubert hearings on ﬁngermark cases

led to the full admissibility of ﬁngermark ev idence in

the courtroom. Judicial notice was given to the fact

that ﬁngerprints are permanent and unique [5].

January 2002 coincides with the ﬁrst decision that

proposes to limit expert testimony on ﬁngerprint iden-

tiﬁcation. Indeed in U.S. v. Llera Plaza (188F. Supp. 2d

549, 572–73 (E.D. Pa. 2002)), the defense ‘‘Motion to

Preclude the United States from Introducing Latent

Fingerprint Identiﬁcation Evidence’’ has been partly

successful. Judge Pollak held that a dactyloscopist

could not give an opinion of identiﬁcation, and re-

quired that the expert limits his testimony to outline

the correspondences observed between the mark and

the print, leaving to the court the assessment of the

signiﬁcance of these ﬁndings. That led the Government

experts to ask for reconsideration bringing to the de-

bate background documents in relation to the move of

the UK toward the abandonment of the 16 point stan-

dard. Judge Pollak later reversed his opinion, and ad-

mitted the evidence.

Two cases of wrongful ﬁngermark identiﬁcation

following the case of the Scottish police ofﬁcer Shirley

McKie perpetuated this controversy. In the ﬁrst case

the Amer ican Stephan Cowans was convicted by ﬁn-

gerprint identi ﬁcation, but later exonerated by DNA

analysis. In the second case, the American Brandon

Mayﬁeld was wrongly associated with the 11 March

2003 Madrid bombing, by means of ﬁngerprint to a

latent mark revealed by the Spanish National Police on

a plastic bag containing detonators recovered from a

stolen van associated with these bombings. Three FBI

experts and an independent court-appointed expert all

identiﬁed Mayﬁeld as the donor of the mark. Mayﬁeld,

a lawyer based in the US State of Oregon, came to the

Fingerprint, Forensic Evidence of

533

FBI’s attention when one of the latent marks sent by

the Spanish authorities through Interpol gave a hit

against his name on the FBI integrated AFIS (IAFIS),

containing about 440 millions of ﬁngerprints from

44 millions of persons. Brandon Mayﬁeld was arrested,

and remained in custody for a few weeks until the

Spanish dactyloscopists, who immediately had raised

issues with this identiﬁcation, ﬁnally identiﬁed the

mark with the ﬁnger of an Algerian suspect.

The FBI offered an apology and published a re-

search report in the beginning of 2004 in which the

existing FBI procedures were investigated extensively.

This report showed that the mistake in this case was

not owed to the methods the FBI used, but was the

consequence of ‘‘human error’’ which cannot be ex-

cluded. The problem with this frequently used expla-

nation is that the method and the human cannot be

separated in case of an activity at which the human acts

as a measuring instrument as is the case in traditional

dactyloscopy [11].

An extensive research by the General Inspector of

the US department of Justic e appeared in January 2006 in

which a clear analysis was given of the facts and circum-

stances causing the incorrect identiﬁcation [12]. Accord-

ing to this report, an important factor in the Mayﬁeld

case was that when a search is performed using a very

large database, there will always be a reference print

which strongly looks like the unknown mark. A posi-

tive consequence of these cases is that they initiated a

move towards a much more open discussion about the

misidentiﬁcations in the forensic ﬁngerprint ﬁeld.

Analysis of the Current Practice

Research in embryology and statistics clearly do not

legitimate the reduction of ﬁngerprint individuality to

counting minutiae. Indeed the scope of features is

much broader than minutiae alone, and the nature of

the papillary individuality prevents the adoption of

any predeﬁned number of ridge characteristic s neces-

sary for identiﬁcation, without signiﬁcant di fferences

[13]. It is axiomatic that no two ﬁngerprints are iden-

tical, as no two entities of any kind can be identical to

each other. A common misconception lies in the fact

that the features of individuality of the ﬁngerprint is

often attributed to the ﬁngermark. As already de-

scribed by Locard, in criminalistics, the transfer of

material is logically never perfect. In dactyloscopy,

the transfer of the pattern from the ﬁngerprint ridges

to the ﬁngermark is accompanied by two types of loss

of information: quantitative, due to the limited size of

the trace, and qualitative, due to di stortion, blurring,

bad resolution, and loss of pore and edge details.

The challenge for dactyloscopy is about the ability

to quantify the information available for the individu-

alization process in a partial distorted ﬁngermark,

and not to prove the individuality of the friction

ridge skin. The ﬁrst step in the quantiﬁcation of the

evidential value of ﬁngermark evidence consists in

estimating the similarity between the features of this

ﬁngermark and those of the ﬁngerprint considered

as potential source of this mark. The second step

consists in estimating the typicality or the rarity of

these features, and the third step, in reporting the

similarity–typicality ratio as evidential value. This con-

cept encapsulates a continuum of values for individu-

alization of the ﬁngermarks ranging from very high to

very low, depending on the feature analyzed. There-

fore, the forensic individualization process of ﬁnger-

marks cannot be considered as a binary decision

process, but has to be envisaged as a purely probabilis-

tic assessment of the value of evidence, as it is for any

type of evidence [14].

Probabilistic models, which are applicable to ﬁn-

germark indiv idualization [15], have been proposed

and accepted by forensic scientists in other forensic

areas – i.e., DNA, microtraces and speaker recognition

[16]. The absence of extensive statistical analysis

on ﬁngerprint variability can be viewed as the main

reason to prevent giving qualiﬁed opinions. Statistical

data only support and comfort identiﬁcation state-

ments used by dactyloscopists but, according to

Stoney, ‘‘we must realize that to reach absolute identi-

ﬁcation, or its probabilistic equivalent, through an

objective process is not possible. Probabilities are ob-

jective when they can be tested and reproduced’’ [17].

Future Perspectives

The statistical studies applied to ﬁngerprints and ﬁn-

germark individualization provide valuable knowledge

about the statistical behav ior of various types of fea-

tures, mainly the minutiae, and to a more limited

extent, the pores, but they do not provide a robust

tool to assess the probability associated with a given

conﬁguration of features for several reasons: none of

the proposed models has been subjected to an extend-

ed empirical validation, and the assumptions about

534

Fingerprint, Forensic Evidence of

the features used in these models have not been fully

explored.

The research possibilities are huge, mainly in three

different directions. The ﬁrst is a reﬁneme nt and an

empirical validation of the model-based approaches

developed in earlier studies [8]. The second is the

development of data-driven approaches taking advan-

tage of the capabilities of the current AFIS systems,

embedding large ﬁngerprint and ﬁngermark databases,

high computation capabilities, and sophisticated pat-

tern recognition techniques. The third direction is to

explore the morphogenesis process from the point of

view of mathematical biology, with the aim to

determine the contribution of the genetic, environ-

mental, and the other factors, which inﬂuence the

features deﬁned in the three levels of informat-

ion present in the ﬁngerprint. These studies require

the availability of large samples of ﬁngermarks

and ﬁngerprints and a clear deﬁnition of the features

used by the examiners to compare ﬁngermarks with

ﬁngerprints.

Related Entries

▶ Automatic Fingerprint Matching

▶ Fingerprint Classiﬁcation

▶ Fingerprint Databases and Evaluation

▶ Fingerprint Features

▶ Fingerprint Individuality

▶ Fingerprint Matching, Manual

▶ Individuality of Fingerprints

▶ Latent Fingerprint Experts

References

1. Meuwly, D.: Forensic individualization from biometric data. Sci.

Justice 46(4), 205–213 (2006)

2. Kuecken, M., Newell, A.C.: Fingerprint formation. J. Theor. Biol.

235, 71–83 (2005)

3. Wertheim, K., Maceo, A.: The critical stage of friction ridge and

pattern formation. J. Forensic Ident. 52(1), 35–85 (2002)

4. Jain, A.K., Prabahakar, S., Pankanti, S.: On the similarity of

Identical Twin Fingerprints. Pattern Recognit. 35(12),

2653–2663 (2002)

5. Champod, C., et al.: Fingerprints and other Ridge Skin impres-

sions. CRC press, London (2004)

6. Ashbaugh, D.R.: Qualitative-quantitative friction ridge analysis –

An introduction to basic and advanced ridgeology. In: Geberth,

V.J. (ed.) Practical Aspects in Criminal and Forensic Investiga-

tions. CRC Press, Boca Raton, FL (1999)

7. Stoney, D.A.: Measurement of ﬁngerprint individuality. In: Lee,

H.C. Gaensslen, R.E. (eds.) Advances in Fingerprint Technology,

pp. 327–388. CRC Press, Boca Raton, FL (2001)

8. Neumann, C., et al.: Computation of likelihood ratios in ﬁnger-

print identiﬁcation for conﬁgurations of any number of minu-

tiae. J. Forensic Sci. 52(1), 54–64 (2007)

9. Berry, J., Stoney, D.A.: History and development of ﬁngerprint-

ing. In: Lee, H.C., Gaensslen, R.E. (eds.) Advances in Fingerprint

Technology, pp. 1–40. CRC Press, Boca Raton, FL (2001)

10. McCabe, R. M. (ed.): Data format for the interchange of ﬁnger-

print, facial, scar mark & tatoo (SMT) Information, American

National Standard ANSI/NIST-ILT 1-2000, July (2000)

11. Fine, G.E.: A Review of the FBI’s handling of the Brandon

Mayﬁeld case. 2006, Ofﬁce of the Inspector General, U.S.

Department of Justice

12. Ofﬁce of the Inspector General, United States Department of

Justice. A Review of the FBIs Handling of the Brandon Mayﬁeld

Case: Unclassiﬁed Executive Summary, Washington, DC (2006)

13. Champod, C.: Dactyloscopy: Standards of proof, In: Siegel, J.

(ed.) 1Encyclopedia of Forensic Science. Academic, London.

(2000)

14. Taroni, F., Champod, C., Margot, P.: Forerunners of Bayesianism

in early forensic science. Jurimetrics 38, 183–200 (1998)

15. Good, I.J.: Weight of evidence and the Bayesian likelihood ratio,

In: Aitken, C.G.G. (ed.) Statistics and the Evaluation of Evi-

dence for Forensic Scientists. Wiley, Chichester, UK (1995)

16. Aitken, C.G.G., Taroni, F.: Statistics and the evaluation of evi-

dence for forensic scientists. Wiley, Chichester, UK (2004)

17. Stoney, D.A.: What made us ever think we could individualize

using statistics. J. Forensic Sci. Soc. 31(2), 197–199 (1991)

Fingerprint, Palmprint, Handprint

and Soleprint Sensor

GEP PY PARZIALE

Cogent Systems, Inc., South Pasadena, CA, USA

Synonyms

Fingerprint device; Fingerprint sensor; Handprint

sensor; Palmprint device; Palmprint sensor; Soleprint

device; Soleprint sensor

Definition

A ﬁngerprint or palmprint or handprint or soleprint

sensor is a transducer that converts the ridge–valley

structure of a person’s hand or foot sole to an electri-

cal signal. Generally, the sensor reads the difference of

Fingerprint, Palmprint, Handprint and Soleprint Sensor

535

pressure, temperature, light, electrical capacit y or other

kinds of energies are measured between the ridges and

the valleys. Then, this difference is converted into an

electrical digital signal that is encoded as an image

representing the ridge–vall ey pattern. Different tech-

nologies can be applied to achieve this conversion and

each of them brings advantages and disadvantages.

It is important to highlight that the output signal is

a representation of the real-world ridge–valley pattern.

Hence, if F is a ridge–valley pattern of a real-world

ﬁnger tip and s is the transfer function of a device, the

output signal is F

¼s(F ) and F

6¼ F.

Introduction

The similarity of the ridge–valley pattern of the epider-

mis present on ﬁnger tips, palms, and soles [1] allows

to use the same physical principles for capturing ﬁn-

gerprints, palmprints and soleprints. The devices using

these technologies can be grouped into a single family,

known as

▶ livescan furrow devices or shortly, livescan

de vices.

The technological advancement of livescan devices

has been mainly driven by the research done in the

ﬁngerprint recognition ﬁeld more than the palmprint

and soleprint modalities. The reason has to be found in

a more convenient use of ﬁngerprint devices, instead of

the larger, heavier, and more power-consumer palm-

print and soleprint ones. Moreover, ﬁngerprint being

the oldest biometric mean s used to identify people,

large collection of data have always been available.

This facilitated the development of algorithms for ﬁn-

gerprint recognition, pushing experts and scientists to

focus mainly on this modality more than palmprints

and soleprints.

Ink-on-Paper Method

The oldest approach to capture the furrow pattern is

represented by the ink-on-paper method. Even if we

cannot consider this as a real sensing technology, it is

important to mention it here, since ink -on-paper is

still widely used to collect palmprints, ﬁngerprints,

handprints, and soleprints. Moreover, it represents a

strong obstacle for the advancement and the introduc-

tion of new capture technologies. The reason has to be

found in the existence of very large databases collected

using this method during the last ten decades. When a

new technology is introduced on the market, it must

have a high degree of interoperability with the ink-on-

paper method to ensure the continuity of the use of

these databases, because a ﬁngerprint or palmprint

representation different than the legacy one would

make the comparison very difﬁcult. Thus, the repre-

sentation of the ridge–valley pattern provided by the

ink-on-paper method still represents the mod el that

the modern technology tries to imitate.

The ink-on-paper capture approach consists in

covering the ridge–valley pattern with black ink.

Then, the print is obtained impressing the inked skin

onto a white paper applying a small pressure. The

resulting print is represented by a black mark for each

ridge, while nothing is left in correspondence of each

valley. The quantity of the ink applied on the skin and

the pressure applied onto the paper during the impres-

sion are very important factors inﬂuencing the quality

of the ﬁnal result. In spite of other approaches, this

technique does not suffer the skin condition problems

(dry skin, wet skin, etc.), which are instead very difﬁcult

to overcome in the case of the other capture methods.

In some applications, the capture of ﬁngerprints is

performed rolling the ﬁnger onto the paper. This

is done to acquire as much information as possible of

the ﬁnger tip that can be used during an identiﬁ cation.

The impression obtained with this approach is called

▶ rolled-equivalent ﬁngerprint. Using dedicated image

processing algorithms, rolled-equivalent ﬁngerprints

can also be obtained rolling the ﬁnger on the sensing

surface of a sensor.

Collecting ﬁngerprints, palmprints, handprints,

and soleprints with the ink-on-paper method is still

widely used, because it still represents the cheapest

way to collect these biometric data. In Spain, the regis-

tration of all new born children is done applying ink on

the baby soles of the feet and impressing them on a

paper. In some Asian countries, inked ﬁngerprints are

used to register civilians during elections to avoid dou-

ble voting. In Switzerland, Spain, Germany, and many

other Countries, criminals are still registered by inking

the tips of their ﬁngers and their hands to collect the ten

ﬂat and rolled ﬁngerprints and palms.

Once the ﬁngerprints are collected on the paper,

they can be digitalized using a ﬂatbed scanner and then

stored in digital format. This approach is still the most

used by an Automated Fingerprint Identiﬁcation Sys-

tem (AFIS) or an Automated Palmprint and Finger-

print Identiﬁcation System (APFIS).

536

Fingerprint, Palmprint, Handprint and Soleprint Sensor

To improve the user convenience, especially in ci-

vilian applications, a special transparent oily substance

is used in place of the black ink. In this way, the

ﬁngerprinted person does not need to wash her/his

hands many times to remove the inconvenient residues

of the black ink.

Nowadays, palmprints are becoming more and

more popular in crime investigation, especially for

latent comparison, since recent studies have demon-

strated that more than 30% of the latent prints found

on a crime scene belongs more likely to palms than

ﬁngers. Generally, the ink-on-paper palm capture

consisted of inking the lower palm and the impress

it on a paper. Nowadays, the reduction of the cost of

the digital storage space allows to store larger quantity

of data. Thus, the most modern approach consists of

capturing the full handprint consisting of the three

▶ palm segments (lower, upper and writer palms).

Sensor Characteristics

Before describing the modern ﬁngerprint sensor tech-

nologies, their main characteristics are highlighted in

this article . These features deﬁne the application range

in which the sensor can be used. For some applications,

livescan devices have to pass very strict tests. The most

famous and required certiﬁcation is the FBI ﬁngerprint

scanner certiﬁcation, covered in the Appendix F of the

Criminal Justice Information Service (CJIS) Electronic

Fingerprint Transmission Speciﬁcation [2]. A list of

FBI certiﬁed livescan devices is available at http://

www.fbi.gov/hq/cjisd/iaﬁs/cert.htm.

The ﬁrst important feature for a livescan device is

the Image Resolution, which describes the ability of a

sensor to distinguish, detect, and/or record physical

details of the ridge–valley pattern. It represents the num-

ber of pixels in a unitary length and is expressed in pixel-

per-inch or shortly, ppi. Typical image resolution values

are 500 and 1,000 ppi. The ﬁrst value is th e most

common and it is used in majority of the applications

and products present in the market. The 1,000 ppi is

mainly used for criminal investigation, especially for

palmprints. The interest in analyzing the so-called

third level details of a ridge–valley pattern is now

pushing the manufac turers to introduce new devices

with very high resolution (1,500–5,00 0 ppi).

Radiometric Resolution or Image Depth or Dynamic

Range dete rmines how ﬁnely a sensor can represent

or distinguish differences of intensity. It is usually

expressed as a number of gray levels or bits, for exam-

ple, 8 bits or 256 gray levels which is typical of ﬁnger-

print image.

The Modulation Transfer Function (MTF) or Spatial

Frequency Response is another important parameter.

Spatial frequency is typically measured in cycles or

line pairs per millimeter (lp∕mm). The more extended

the response, the ﬁner the detail and the sharper the

image. MTF is the contrast at a given spatial frequency

f relative to contrast at low frequencies and it can be

computed with the following Eq. (1):

MTF ¼ 100%

Cðf Þ

Cð0Þ

; ð1Þ

where C(f ) ¼ (V

max

 V

min

)∕(V

max

þ V

min

) is the con-

trast at frequency f, and C(0) ¼ (V

 V

)∕(V

þ V

)

is the low frequency contrast. V

, V

min

and V

max

represent the luminance for black areas, the luminance

for white areas, the mini mum luminance for a pattern

near spatial frequency f and the maximum luminance

for a pattern near spatial frequency f respectively.

All the optical features of a sensor can be measured

using special targets. To test the quality of a device,

a manufacturer must purchase these targets and test

the accuracy of all its optical features.

The Geometric Image Accuracy represents the abso-

lute value of the difference D ¼ XY, between the

distance X measured between any two points on the

target and the distance Y measured between those same

two points on the output image. This is a very impor-

tant parameter especially for devices having a very

large capture area.

The Grayscale Linearity (GL) represents the capacity

of a device to reproduce the gray level values cor-

rectly. A target with gradually vary ing grayscale levels

is used for this scope. The grayscale levels on the

output image are compared with the grayscale levels

on the input target to measure the accuracy of the

representat ion.

The Signal-to-Noise Ratio (SNR) is measured using

another special target representing a grayscale level

reference. This reference can be a white-colored and a

black-colored target. An image is generated from these

two targets and compared po int by point with the

reference.

The Framerate represents the number of frames a

sensor can generate per time unit. It is measured in

frames ∕s and it is a very important parameter when the

Fingerprint, Palmprint, Handprint and Soleprint Sensor

537

object (ﬁnger, palm or hand) movements are implied

during a capture (sweep devices). To improve the ﬁnal

image quality, many sensors acquire more images of the

same ﬁnger or palm during an acquisition. The captured

images are then combined to produce the ﬁnal image.

The Shutter-speed is the time that a detector needs

to capture a single image.

Other important sensor characteristics that can

change the application range of a device are the commu-

nication interface type (USB, Firewire, Ethernet, etc.),

the sensor dimensions and weight, the Mean-Time-

Before-Failure (MTBF), the self-powering capacity (if

the power of the device comes from the communica-

tion interface) and obviously the price.

Optical Sensors

Sensors using light to discriminate between ridges and

valleys represent the oldest technology to capture ﬁn-

gerprints, palmprints, and soleprints with no need of

inking hands and feet.

The most widely used optical capture principle is

known as Frustrated Total Internal Reﬂection (FTIR)

and highlighted in Fig. 1. The sensor contains an opti-

cal prism and one of its faces is used as the

▶ platen

that must be touched by the ﬁnger to produce an

image. A monochromatic light source enters the

prism and is reﬂected in accordance with each valley

of the skin. Then, it is collected by a detector (CM OS

or CCD array). The light is absorbed in accordance

with each ridge touching the platen. The lack of reﬂec-

tion allows the ridges (appearing dark in the image) to

be discriminated by the valleys (appearing bright in the

image). The 3D ridge–valley structure plays here an

important role: presenting to the platen a picture or a

drawing of a ﬁngerprint does not produce any image.

On the other hand, molding the shape of the ridge–

valley pattern with special materials (latex, silicon, etc.)

and touching the platen with it produces an image that

cannot be distinguished by the image obtained by the

real ridge–valley pattern (spooﬁng).

This capture technology is strongly inﬂuenced by

the skin conditions. When the skin is too dry, the ridges

do not completely adhere to the glass platen and thus,

an image with very low contrast is obtained. On the

other hand, very wet ﬁngers produce an uniform black

spot image, because during the ﬁnger pressure, the

sweat accumulates in accordance with each valley

and the light of the LEDs is fully absorbed. The result

is a uni form dark spot with very low contrast between

ridges and valleys. To overcome these problems, before

each capture the user is asked to clean her/his hands

and wet them with special no n-toxic substance pro-

viding the right quantity of wet on the skin.

Another problem related to optical devices is repre-

sented by the so called

▶ halo effect. When the wet

skin touches the colder platen, the moisture starts to

Fingerprint, Palmprint, Handprint and Soleprint Sensor. Figure 1 Total frustrated internal reflection (▶ TFIR) principle.

538

Fingerprint, Palmprint, Handprint and Soleprint Sensor

condense. This results in a halo on the ﬁnal image

reducing the ridge–valley contrast. To avoid the halo

effect, the platen has to be warmed up. This process is

expensive in terms of current consumption especially

for palmprint and soleprint devices with very large

platen. The warming process requires a certain period

of time and thus, these devices cannot be used imme-

diately after they are switched on. This is sometimes

impractical for some applications.

The physical size is another limitation of the optical

sensors. The length of the optical path (the path tra-

versed by the light from its source to the detector)

cannot be signiﬁcantly reduced without introducing

severe optical distortions on the ﬁnal image. The use

of small mirrors and special lenses can help in keeping

the same path length in a small space, but the

manufacturing costs drastically increase and the ro-

bustness of the device decreases.

Nowadays, optical sensors represent the maturest

technolog y in the market for capturing ﬁngerprints,

palms and soleprints. The large production of

this kind of devices is reducing their price more and

more. Since they are rugged and less sensitive to

environmental factors than other technologies, optical

sensors are spreading very fast and blocking the pene-

tration of other capture technologies into the market.

Palmprint devices,

▶ slap or four-four-two devices

and

▶ Rolls Capture Devices are only available based

on this technology for the qualit y it can provide also in

the case of devices with large platens. This is why

palmprint sensors are only available based on optical

technolog y.

Optical Multispectral Sensors

To improve the ridge detail representation provided by

the FTIR method, a novel approach to capture ﬁnger-

print has been recently proposed [3–5]. This approach

is called Multispectral Imaging (Fig. 2) and uses mul-

tiple illumination wavelengths rather than a single

monochromatic illumination commonly used in the

FTIR approach. The orthogonal conﬁguration of linear

polarizers emphasizes this multispectral light, which

penetrates the surface of the skin. The light then under-

goes multiple scattering events before emerging from

the skin toward the image array. In avoiding the

optical phenomenon of the FTIR, the multispectral im-

aging sensor is capable of collecting more identifying

data from the ﬁnger than the FTIR sensor. Currently,

only ﬁngerprint devices are available based on this

technology.

Optical Multispectral Imaging is also claimed to be

capable of detecting fake ﬁngers obtained with organic

or synthetic materials. The difference between the

spectral characteristics of the skin and these materials

is known and can be used to detect fake ﬁngerprint.

Optical Contactless or Touchless

Sensors

When a ﬁnger touches or rolls onto a surface, the elastic

skin deforms. The quantity and direction of the pres-

sure applied by the user, the skin conditions, and the

projection of an irregular 3D object (the ﬁnger) onto a

Fingerprint, Palmprint, Handprint and Soleprint Sensor. Figure 2 Multispectral imaging principle.

Fingerprint, Palmprint, Handprint and Soleprint Sensor

539