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

Подождите немного. Документ загружается.

Definition

SFinGe (Synthetic Fingerprint Generator) is a ﬁnger-

print sample synthesis approach developed by the

Biometric System Laboratory of the University of

Bologna (Italy). It is available as a software program

able to generate large databases of images very similar

to human’s ﬁngerprints, together with

▶ ground-truth

data about their characteristics and features. These

databases are particularly useful for developing, opti-

mizing and testing ﬁngerprint recognition systems and

are being extensively used by industrial, academic and

government organizations.

Overview and History

SFinGe (the Italian for sphinx, pronunciation sphin-je)

is the acronym for Synthetic Fingerprint Generator.

SFinGe can be used to easily create large databases of

ﬁngerprints, thus allowing recognition algorithms to

be simply trained, tested and optimized. The images

generated emulate ﬁngerprints acquired with on-line

sensors (see

▶ Fingerprint Acquisition) but, with a few

changes, the simulation of impressions produced by

the ink-technique is also possible.

This

▶ ﬁngerprint sample synthesis approach

was developed at the Biometric System Laboratory of

the University of Bologna (Italy) [1] since 1999; the

ﬁrst version of the method [2] was able to synthesize

realistic ﬁngerprint patterns, but limited to only one

impression of each ‘‘synthetic ﬁng er’’. Appropriate

techniques for simulating more impressions of the

same ﬁnger were developed at the end of 2000 , suc-

cessfully adopted to generate one of the test databases

for the ﬁrst Fingerprint Veriﬁcation Competition [3],

and described, for the ﬁrst time, in [4]. In 2002, realis-

tic background generation capabilities were added [5]

and, in 2004, an improved noise model was developed

[6]. More recently, the approach has been expanded

with the generation, for each synthetic ﬁngerprint

image, of

▶ ground-truth minutiae information

(i.e., the precise location and characteristics of each

minutia) and other features (such as the

▶ orientation

ﬁeld)[7].

A software tool for generating ﬁngerprint images

according to the SFinGe method has been provided

by the Biometric System Laboratory of the University

of Bologna since 2001. A demo version of this tool

(Fig. 1) can be downloaded from http://biolab.csr.

unibo.it/sﬁnger.htm l and has been used to prepare

most of the ﬁgures in the following sections.

The Generation Process

Fig. 2 shows a functional schema of the generation

approach. SFinGe adopts a statistical ridge pattern

model (see ﬁngerprint sample synthesis) to create

a master-ﬁngerprint (that is the unique and immutable

SFinGe. Figure 1 The user interface of the SFinGe software tool.

1170

SFinGe

characteristics of a ‘‘synthetic ﬁnger’’) through the

following steps [5]:

1. Fingerprint shape generation: Deﬁnition of the global

shape of the ﬁngerprint, according to a simple

model based on elliptical segments;

2. Orientation ﬁeld generation: A mathematical ridge-

ﬂow model allows to generate a consistent orienta-

tion ﬁeld;

3. Frequency map generation: The local ridge-line fre-

quency is generated on the basis of some heuristic

criteria;

4. Ridge-line pattern generation: Ridge-lines and min-

utiae are created using space-variant ﬁltering.

Once a master-ﬁngerprint has been created, one ore

more of its ‘‘impressions’’ can be randomly generated,

by applying the following steps [5]:

5. Selection of the contact region: The ridge-line pat-

tern is translated without modifying the global

ﬁngerprint shape and position (this simulates

different ﬁnger placements over the acquisition

device);

6. Variation of the ridge-line thickness: Morphological

operators are applied to simulated different degrees

of skin dampness and/or ﬁnger pressure;

7. Fingerprint distorti on: A skin disto rtion model is

adopted to simulate the effects of skin elasticity;

8. Noising and rendering: A gray-scale noisy image

is produced by modeling some of the factors that

deteriorate the quality of real ﬁngerprints;

9. Global translation/rotation: The image is randomly

translated and/or rotated, to simulate real ﬁnger-

prints that usually are not perfectly centered and

can present a certain amount of rotation;

10. Background generation: A realistic background

can be created to simulate a given acquisition

device.

Fig. 2 shows, for each of the steps described above, the

various input parameters (graphically represented by

the red arrows). The SFinGe software tool lets the user

SFinGe. Figure 2 A functional schema of the SFinGe generation approach: each rounded-box represents a step (based

on a corresponding mathematical model); the main parameters are reported between square brackets. Steps 1–4

create a master-fingerprint, steps 5–10 generate the final synthetic image.

SFinGe

1171

adjust most of those parameters and observe the

corresponding effects on the synthetic ﬁngerprint (see

Figs. 3–7). The software also allows a database of syn-

thetic ﬁngerprints to be batch-generated, given a relatively

small set of input parameters (see Fig. 8), including:

number of ﬁngers, impressions per ﬁnger, image size,

seed for the random number generator, maximum

amount of noise, maximum amount of deformation.

During the batch-generation of a ﬁngerprint database,

each master-ﬁnger is gen erated by using a different

seed for the random number generator; those seeds

are randomly selected as well. During the generation

of a single database, all the seeds chosen are different;

although it is reasonable to assume that different seeds

imply different ﬁngerprints, it might happen that two

different seeds produce almost identical ﬁnge rprint

images. To reduce this hypothetical risk, SFinGe adopts

one of the best

▶ pseudo-random number generators

proposed in the scientiﬁc literature [8].

The creation of minutiae ground-truth proceeds

in parallel with the ﬁngerprint generation (Fig. 9): the

standard minutiae extraction procedure deﬁned in [9]

is applied to the master-ﬁngerprint, then all the rele-

vant transformations executed on the ﬁngerprint are

applied to the minutiae (e.g., translation, rotation,

distortion). This approach has some clear advantages:

1. The features can be extracted by applying the stan-

dard procedures easily and without ambiguities,

since the extraction occurs on a binary image with-

out any noise;

2. The ground truth is always unique and sound,

even when the quality of the ﬁnal image is very

low (see Fig. 10).

SFinGe. Figure 3 Graphical user interface for the fingerprint shape generation (left) and orientation field generation

(middle and right).

SFinGe. Figure 4 Graphical user interface for the ridge-line pattern generation.

1172

SFinGe

SFinGe. Figure 6 Graphical user interface for fingerprint distortion (left) and noising (middle and right).

SFinGe. Figure 5 Graphical user interface for the variation of the ridge-line thickness.

SFinGe. Figure 7 Graphical user interface for global rotation/translation (left) and background generation (middle

and right).

SFinGe

1173

Generation of other ground-truth features is per-

formed in a similar fashion; for instance, all the rele-

vant transformations can be applied to the orientation

ﬁeld calculate d at step 2, thus obtaining the true orien-

tation ﬁeld of the ﬁnal synthetic ﬁngerprint.

The automatic generation of a whole ﬁngerprint

database (including ground-truth data) is totally

parallelizable, since the generation of each master-

ﬁngerprint (with its impressions) is independent of

the others. This makes it possible to distribute the

SFinGe. Figure 8 The batch-generation options.

SFinGe. Figure 9 Generation of ground-truth minutiae data.

1174

SFinGe

process on many computers; for instance, using ten

3GHz PCs in a network, a database of 100,000 ﬁnger-

prints (10,000 ﬁngers, 10 impressions per ﬁnger) can

be generated in less than two hours.

Applications

SFinGe can be used to create, at zero cost at without

any

▶ privacy issue, large databases of ﬁngerprints,

whose main characteristics can be controlled and

adjusted according to the speciﬁc needs of a given

application. Furthermore, ground-truth data about

the main ﬁngerprint features can be automatically

produced for each ﬁngerprint in the database. SFinGe

has been used by many industr ial, academic and gov-

ernment organizations; the main applications of this

synthesis approach are described in the following .

1. Per formance evaluation: SFinGe is an effective tool to

over come the problem of collecting large ﬁngerprint

databases for test purposes. Obviously real ﬁnger-

print databases cannot be completely substituted,

especially when performance has to be measured

referring to a given real environment/application;

on the other hand synthetic ﬁngerprints proved to

be well suited for

▶ technology eva luations [10]: a

comparison of the behavior of several ﬁngerprint

matching algorithms on real and synthetic data-

bases showed that not only the performance is

very similar, but the genuine/impostor distribu-

tions and the FMR/FNMR curves (see

▶ Perfor-

mance Measures) are also surprisingly close [5].

2. Training: Many classiﬁers and pattern recognition

techniques (e.g., neural networks,

▶ Principal

Component Analysis,

▶ Support Vector Machines )

require a large training set for an accurate learni ng

stage. Synthetic ﬁngerprint images are very well

suited to this purpose: in fact the generator input

parameters allow to explicitly control the type and

features of the generated sets (e.g ., class, type of

noise, distortion) and this can be exploited in con-

junction with boosting techniques to drive the

learning process. For example, in [11], a large syn-

thetic training set (generated by SFinGe) was suc-

cessfully used to derive optimal parameters for

▶ ﬁngerprint indexing.

3. Security evaluation : Synthetic ﬁngerprints can be

used to test the robustness of ﬁngerprint veriﬁca-

tion systems to ‘‘Trojan horse’’ attacks against the

sensor or the feature extractor [5] (see

▶ Biometric

Security, Overview). SFinGe allows to generate large

sets of ﬁngerprints whose features (e.g. minutiae

distribution) can be varied independently of other

ﬁngerprint characteristics (e.g. orientation ﬁeld),

and therefore, it is well suited for studying the ro-

bustness against ‘‘hill-climbing’’ attacks (see [5]).

4. Semantic conformance to standards: Interoperability

tests [12, 13] have shown that the location, direc-

tion, and type of minutiae extracted by different

minutiae extraction algorithms from the same ﬁn-

ger image tend to be different (see

▶ Finger Data

Interchange Format, Standardization). Algorithms

syntactically compliant to standards such as the

ISO/IEC 19794-2 [9], are often not semantically

compliant and this creates huge interoperability

problems. Unfortunately, testing semantic confor-

mance to a minutiae extraction standard is not easy,

since it requires a lot of data with manually-labeled

minutiae points (ground-truth); furthermore, in low-

quality areas, even the manual labeling of minutiae

points is not reliable. The automatic generation of

ground-truth data for synthetic ﬁngerprint images

provided by SFinGe is an effective way to carry out

semantic conformance and interoperability studies.

For instance, in [7] a synthetic database has been

used to analyze the distribution of minutiae posi-

tions and directions of some algorithms with re-

spect to the ground-truth (see Fig. 11).

SFinGe. Figure 10 Minutiae ground-truth as generated by SFinGe for very high-quality fingerprints (left),

medium-quality fingerprints (middle), and low-quality fingerprints (right).

SFinGe

1175

Related Entries

▶ Anatomy of Fingerprint

▶ Finger Data Interchange Format, Standardization

▶ Fingerprint Classiﬁcation

▶ Fingerprint Databases and Evaluation

▶ Fingerprint Features

▶ Fingerprint Indexing

▶ Fingerprint Noise

▶ Fingerprint Orientation Field

▶ Fingerprint Ridge-line Pattern

▶ Fingerprint Sample Synthesis

▶ Fingerprint Singularities, Minutiae, Pores

▶ Performance Measures

▶ Privacy Issues

▶ Support Vector Machine

References

1. Biometric System Laboratory (University of Bologna) web site,

Available at: http://biolab.csr.unibo.it. Accessed January 2008

2. Cappelli, R., Erol, A., Maio, D., Maltoni, D: ‘‘Synthetic

Fingerprint-image Generation.’’ In: Proceedings of the 15th In-

ternational Conference on Pattern Recognition, Barcelona,

vol. 3, pp. 475–478 September (2000)

3. Maio, D., Maltoni, D., Cappelli, R., Wayman, J.L., Jain, A.K.:

FVC2000: Fingerprint Veriﬁcation Competition. IEEE Trans

Pattern Anal Mach Intell 24(3), 402–412 (2002)

4. Cappelli, R., Maio, D., Maltoni, D: ‘‘Synthetic Fingerprint-

Database Generation.’’ In: Proceedings of the 16th International

Conference on Pattern Recognition, Que

bec City, vol. 3,

pp. 744–747, August (2002)

5. Cappelli, R: Synthetic ﬁngerprint generation, In: Maltoni, D.,

Maio, D., Jain, A.K., Prabhakar, S. (eds.) Handbook of Finger-

print Recognition, Springer, New York, (2003)

6. Cappelli, R., Maio, D., Maltoni, D., ‘‘An Improved Noise Model for

the Generation of Synthetic Fingerprints.’’ In: Proceedings Eighth

International Conference on Control, Automation, Robotics and

Vision (ICARCV2004), Kunming, China, December (2004)

7. Cappelli, R: ‘‘Use of Synthetic Data for Evaluating the Quality of

Minutia Extraction Algorithms’’, In: Proceedings of the Second

NIST Biometric Quality Workshop, Gaithersburg, Maryland,

November (2007)

8. Saito, M., Matsumoto, M: ‘‘SIMD-oriented Fast Mersenne

Twister: a 128-bit Pseudorandom Number Generator’’, Monte

Carlo and Quasi-Monte Carlo Methods 2006, pp. 607–622.

Springer, New York, (2008)

9. ISO/IEC 19794-2:2005, Information technology – Biometric

data interchange formats – Part 2: Finger minutiae data

10. Cappelli, R., Maio, D., Maltoni, D., Wayman, J.L., Jain, A.K.

Performance Evaluation of Fingerprint Veriﬁcation Systems.

IEEE Trans Pattern Anal Mach Intell. 28(1), 3–18 (2006)

11. Cappelli, R., Maio, D., Maltoni, D: ‘‘Indexing Fingerprint

Databases for Efﬁcient 1:N Matching.’’ In: Proceedings of the

Sixth International Conference on Control, Automation, Robot-

ics and Vision (ICARCV2000), Singapore, December (2000)

12. Minutiae Interoperability Exchange Test (MINEX) web site,

Available at: http://ﬁngerprint.nist.gov/minex04. Accessed

January 2008

13. Minutiae Template Interoperability Testing (MTIT) Project

web site, Available at: http://www.mtitproject.com. Accessed

January 2008

Shape

‘‘Shape is all the geometric information that remains

when location, scale and rotational effects are ﬁltered

out from the object’’. Kendall’s statistical shape is

a sparse descriptor of the shape that describes

SFinGe. Figure 11 Distributions of minutiae placement and direction as estimated in [7] for some feature extraction

algorithms. In each image, the intensity I[x, y] is proportional to the estimated likelihood that a minutia will be found by an

algorithm at position (x, y) with respect to the ground-truth minutia direction (denoted by the arrow).

1176

Shape

the shape conﬁguration of k landmark points in an m-

dimensional space as a k  m matrix containing the

coordinates of the landmarks from which scale and

translation are ﬁltered out. This shape feature therefore

lies on a spherical manifold which is well studied and

therefore properly deﬁned distance measures are avail-

able in this manifold to perform recognition.

▶ Gait Biometrics, Overview

▶ Gait Recognition, Model-Based

Shape Index

Shape index Si, a quantitative measure of the shape of a

surface at a point p, is deﬁned by (1),

ðpÞ¼



tan

1

ðpÞþk

ðpÞ

ðpÞk

ðpÞ

ð1Þ

where k1 and k2 are maximum and minimum princi-

pal curvatures, respectively. With this deﬁnition, all

shapes can be mapped into the interval S

2[0,1]. The

larger shape index values represent convex surfaces and

smaller shape index values represent concave surfaces.

▶ Ear Biometrics, 3D

Shape Model

Shape model is often embedded within a structural

model. It uses geometrical shapes e.g., stick ﬁgures as

a reduced representation of human body, blob, or

cylinder to represent body masses, or silhouette outline

of a human ﬁgure obtained from edge information to

describe the body of interest.

▶ Gait Recognition, Model-Based

Shape vs. Texture

▶ Face Recognition, Geometric vs. Appearance-Based

Shielding Functions

A theoretical approach to cancelable biometrics devel-

oped by Linnartz et al. (2003). Shielding functions allow a

veriﬁer to check the authenticity of a prov er (user want-

ing to be veriﬁed) without learning any biometric infor-

mation. The scheme depends on proposed d-contracting

and e-revealing functions, which allow testing whether

measured features are within a range. Under some

assumptions, a biometric may be tested without learning

anything about the biometric feature values. One limita-

tion of this scheme is that biometric samples are assumed

to be perfectly registered. This scheme offers an interest-

ing cryptographic basis for the construction of encrypted

biometric systems, and it is used by several authors.

▶ Cancelable Biometrics

Shoeprint Matching

▶ Footwear Recognition

Shot Noise

For image sensors, photon conversion into electrons at

a pixel is a random event. Random ﬂuctuations in the

number of electrons for a ﬁxed numbe r of photons are

referred to as shot noise. The standard model for shot

noise is that the noise is proportional to the square

root of the number of electrons generated. Hence, the

signal/noise ratio is proportional to √ N, where N is the

number of electrons.

▶ Biometric Sensor and Device, Overview

▶ Iris Device

Side-Channel Attacks

Side-channel attack is an attack against a cryptographic

or biometric security system based on measurements

Side-Channel Attacks

1177

of the implementation of the system, rather than

on weaknesses in the algorithms. For example, side-

channel attacks may use timing, power consumption,

or electromagnetic measurements on the security

device. Side-channel attacks are primarily of concern

for biometric encryption systems and match-on-card

devices where an attack could potentially be mounted

by iteratively improving the presented biometric. Very

little research has been done to explore the feasibil-

ity of side-channel attacks, but the success of attacks

on biome tric template security and biometric encryp-

tion suggests that such attacks are certainly feasible.

▶ Security and Liveness, Overview

Signal to Noise Ratio

Information is transmitted or recorded by variations in a

physical quantity. For any information storage or trans-

mission system, there will be intended variations in the

physical quantity – signal – and unintended variations –

noise. In an analog telephone system, the signal (voice) is

represented by variation in a voltage level. As the signal is

transmitted along a phone line, it can pickup other

unintended variations – e.g., leakage of other signals,

static from electrical storms – that are noise. The ratio

of the signal level to the noise level is the signal to noise

ratio (SNR). Since it is a ratio, SNR is dimensionless.

However, SNR can be expressed as either an amplitude

ratio (voltage ratio for the phone example) or a power

ratio (milliWatts for the phone example). This leads to

confusion. SNR is frequently expressed as the log (base

10) of the ratio. When expressed as a log, the dimension-

less unit of SNR is decibel (dB).

What is signal and what is noise can depend on the

circumstances. Radio waves from lighting are noise to

an AM radio broadcast, but can be signal to a meteo-

rological experiment.

▶ Iris Device

Signature Benchmark

▶ Signature Databases and Evaluation

Signature Characteristics

▶ Signature Features

Signature Corporate

▶ Signature Databases and Evaluation

Signature Databases and Evaluation

MARCOS MARTINEZ-DIAZ,JULIAN FIERREZ

Biometric Recognition Group - ATVS,

Escuela Politecnica Superior, Universidad

Autonoma de Madrid, Campus de Canto blanco,

Madrid, Spain

Synonyms

Signature benchmark; Signature corpora; Signature

data set

Definition

Signature datab ases are structured sets of collected

signatures from a group of individuals that are used

either for evaluation of recognition algorithms or as

part of an operational system.

Signature databases for evaluation purposes are, in

general, collections of signatures acquired using a

digitizing device such as a pen tablet or a touch-

screen. Publi cly available databases all ow a fair

performance comparison of signature recognition

algorithms propose d by independent entities. More-

over, signature databases play a central role in public

performance evaluations, which compare different

recognition algo rithms by using a common experi-

me ntal frame work. This type of databases is covered

in this entr y.

On the other hand, signature databases can also be

a module of a veriﬁcation or identiﬁcation system.

1178

Signal to Noise Ratio

They store signature data and other personal informa-

tion of the enrolled users. This signature database is

used during the operation of the recognition system

to retrieve the enrolled data needed to perform the

biometric matching. This kind of databases is not

addressed here.

Dynamic Signature Databases

Until the beginning of this century, research on auto-

matic signature veriﬁcation had been carried out using

privately collected databases, since no public ones were

available. This fact limits the possibilities to compare

the performance of different systems presented in the

literature, which may have been tuned to speciﬁc cap-

ture conditions. Additionally, the usag e of small data

sets reduces the statistical relevance of experiments.

The lack of publicly available databases has also been

motivated by privacy and legal issues , although the

data in these databases are detached from any personal

information. The impact of the signature structural

differences among cultures must also be taken into

account when considering experimental results on a

speciﬁc database. As an example, in Europe, signatures

are usually formed by a fast writing followed by a

ﬂourish, while in North America, they usually corre-

spond to the signers name with no ﬂourish. On the

other hand, signatures in Asia are commonly formed

by Asian characters, which are composed of a larger

number of short strokes compared with European or

North American signatures.

While some authors have made public the data-

bases used for their experimental results [1], most

current dynamic signature databases are collected by

the joint effort of different research institutions. These

databases are, in general, freely available or can be

obtained at a reduced cost. Many signature databases

are part of larger multimodal biometric databases,

which include other traits such as ﬁngerprint or face

data. This is done for two main reasons: the research

interest on multimodal algorithms and the low effort

required to incorporate the collection of other biomet-

ric traits once a database acquisition campaign has

been organized.

Two main modalities in signature recognition exist.

Off-line systems use signature images that have been

previously captured with a scanner or camera. On the

other hand, on-line systems employ digitized signals

from the signature dynamics such as the pen position

or pressure. These signals must be captu red with spe-

ciﬁc devices such as

▶ pen tablets or ▶ touch-screens.

The most popular databases in the biometric research

community are oriented to on-line veriﬁcation,

although in some of them, the scanned signature

images are also available [2, 3]. Some efforts have been

carried out in the handwriting community to collect

large off-line signature databases such as the GPDS-960

Corpus [4].

Unlike other biometric traits, signatures can be

forged with relative ease. Signature veriﬁcation systems

must not only discriminate traits from different sub-

jects (such as ﬁngerprints) but also must discriminate

between genuine signatures and forgeries. In general,

signature databases provide a number of forgeries for

the signatures of each user. The accuracy of the for-

geries depends on the acquisition protocol, the skill of

the forgers, and on how much time the forgers are let

to train. Nevertheless, forgeries in signature databases

are usually performed by subjects with no prior expe-

rience in forging signatures, this being a limitation to

the quality of fo rgeries.

Most on-line signature databases have been cap-

tured with

▶ digitizing tablets. These tablets are, in

general, based on an electromagnetic principle, allow-

ing the detection of the pen position (x,y), inclination

angles (y,g)¼(azimuth, altitude), and pressure p. They

allow recording the pen dynamics even when the pen is

not in contact with the signing surface (i.e., during

pen-ups). On the other hand, datab ases captured

with other devices such as touch-screens (e.g., PDAs)

provide only pen position information, which is

recorded exclusively when the pen is in contact with

the device surface.

In the following, a brief description of the most

relevant available on-line signature databases is given

in chronological order.

PHILIPS Database

Signatures from 51 users were captured using a Philips

Advanced Interactive Display (PAID) digitizing tablet

at a sampli ng rate of 200 Hz [5]. The following signals

were captured: position coordinates, pressure, azi-

muth, and altitude.

Each user contributed 30 genuine signatures, leading

to 1,530 genuine signatures. Three types of forgeries are

present in the database: 1,470 over-the-shoulder for-

geries, 1,530 home-improved, and 240 professional

Signature Databases and Evaluation

1179