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

Подождите немного. Документ загружается.

knowledge of footwear marks and their etiology is

required. However, some efforts have been made to

automate the identiﬁcation of outsole patterns and

their associated shoe model based on class character-

istics. While this task is feasible, the task of automati-

cally verifying an outsole-to-mark or a mark-to-mark

based on class and individual characteristics is much

more difﬁcult. Current nation-wide footwear data-

bases in the UK comprise over 15,000 shoe models

and are constantly expanding as manufacturers intro-

duce new styles. Crime scene marks are recovered by

diverse techniques and the resulting impressions are

often of poor quality, confounded by details of the

underlying surface and may only represent a partial

impression of the entire outsole. Such factors will make

the fully automatic identiﬁcation or veriﬁcation

of outsoles marks a very difﬁcult, if not impossible,

option.

Automatic matching of footwear patterns has not

been repor ted much in the literature. Early works [4]

have employed semi-automatic methods of

▶ manual-

ly annotated footwear print descriptions using a code-

book of shape and pattern primitives, for example ,

wavy lines, geometric shapes, and logos. Searching for

an example impression then requires its encoding in a

similar manner to that used for the reference database.

This process is laborious and can be the source of poor

performance as similar patterns may be inconsistently

encoded by different users. It is still, however, pre-

dominantly used by Police Forces across the UK and

elsewhere. Two major factors are attributed to this;

ﬁrstly, the ease of und erstanding the coding me tho-

dology and its primitives (i.e., their visual intuitive-

ness), and secondly, a lack of proven and accepted

automated systems based on rigorous standards and

robust performance.

One early work [5] employed an intuitive coding

scheme based on shapes automatically generated from

footwear images using various

▶ image morphology

operators. The spatial positioning and frequencies

of these shapes were used for classiﬁcation with a

neural network. Unfortunately, the authors did not

report any performance statistics for their system.

Later works do not follow this approach but instead

use template matching approaches based on the pat-

tern representation in a suitable transform space. Frac-

tal representations were used in [6, 7] with a mean

square noise error method for classiﬁcation. They

report good results; however the dataset used was

small and contained no spatial or rotational variations.

In [8], Fourier Transforms (FT) are used for the classi-

ﬁcation of full and partial marks of varying quality.

The FT provides a degree of invariance to translations

and encodes spatial frequency information. By incor-

porating duplicate rotated templates a degree of

rotation invariance was also possible. Their approach

was weak on ﬁrst rank precision, and this may have

been due to the large variation in print quality.

Also, the footwear prints were processed globally and

hence noise in the images could have hindered the

quality of useful encoded local information evident

in the print. Despite these failings this approach is

promising and shows the importance of encoding

local information.

In [9] local image features (LIF) were used to make

one-to-one comparisons of patterns in a database of

footwear marks which had been subjected to added

noise and transformed in various ways. This work com-

pared a number of the aforementioned approaches, for

which LIF performed very well, indicating that local

image information is crucial to good matching perfor-

mance especially when considering partial marks. How-

ever as test samples were generated from the training set

it is not clear how this template matching approach

would perform on untrained test samples or in a one-

to-many and pattern search scenario.

Moving away from a template-based approach,

work in [10] employed a histogram approach by quan-

tizing edge directions into 5



intervals. In order to cope

with rotational variations of the image, and hence

translation of the histogram bins, an FT was also ap-

plied. Their approach is useful as now the pattern

information is described in a compact way based on

the content of edges, however, due to histogram nor-

malization and the use of the FT, this approach is not

effective with partial marks.

The ﬂexibility of histogram-based encoding is ef-

fective in many image retrieval and indexing tasks as

has been demonstrated in [11]. It is interesting that the

manual coding schemes still used for footwear impres-

sions bear a strong resemblance to basic histogram

coding methods. The work proposed by [12] proceeds

by encoding footwear patterns into a compact vector

space model using a feature-rich codebook of local

feature descriptions. The codebook is a quantized set

of features derived by applying such feature detectors

560

Footwear Recognition

as maximally stable extremal regions (MSERs), fol-

lowed by the use of robust feature descriptors.

A query shoe impression can be coded using this code-

book and compared against others. This approach

is fully automated and yet still preserves some simi-

larity to manu al coding approaches in that the coded

features resemble the use of annotated features in

the semi-automated systems described previously.

Footwear Recognition. Figure 3 Outsole marks captured using the Printscan method showing different stages

of wear and print quality. From left to right the pattern shown has increasing wear. Note the change in tread pattern

appearance.

Footwear Recognition. Figure 4 Various footwear marks collected from scenes of crime. From left to right, a print

in blood on soft tiles, outsole imprint in mud, an electrostatic lift of dirt from carpet, a casting, and a gel lift of a

dusty mark.

Footwear Recognition

561

This approach is able to cope with rotational changes

and queries in the form of partial impressions. Though

good performance is reported in this study (e.g., for a

reference set of 374 different footwear models, a preci-

sion at ﬁrst rank of 87% was obtained), it is difﬁcult to

compare the performance of differing approaches as

no common datasets have been employed nor is there

agreement on testing methodologies.

A number of issues are not addressed by these meth-

ods. One is the problem of dealing with large appearance

changes in footwear marks as the outsole becomes worn

over time and where the mark is strongly degraded or

partially missing (see Fig. 3). Some work by Su et al. in

[13, 14] looked at methods for assessing and improving

image quality of footwear marks collected using the

Printscan process. Even if good quality marks are ob-

tainable there is still work to be done on how to compare

marks obtained by different methods, for example be-

tween a casting and a gel lift (see Fig. 4). Additionally, a

lack of standards for the digital capture of marks and

the absence of an agreed and openly available dataset

are issues that still need to be addressed.

Summary

The use of footwear impressions both recovered from

crime scenes and acquired from suspect’s shoes have a

signiﬁcant role to play as a ‘‘near-biometric’’ in forensic

investigations. It has, despite its long history as an im-

portant tool of forensics, remained until recently largely

forgotten. This will undoubtedly change in the near

future with a much greater development and application

of mainstream biometric tools and methodologies.

Related Entries

▶ Forensic Applications, Overview

▶ Forensic Barefoot Comparison

References

1. Bodziak, W.J., ed.: Footwear Impression Evidence. CRC Press,

Boca Raton (2000)

2. Parliament: Serious Organised Crime and Police Act 2005,

Elizabeth II. The Stationery Ofﬁce (2005)

3. Hilderbrand, D.S., ed.: Footwear, The Missed Evidence. Staggs

Publishing (1999)

4. Mikkonen, S., Astikainen, T.: Databased classiﬁcation system for

shoe sole patterns-identiﬁcation of partial footwear impres-

sion found at a scene of crime. J. Forensic Sci. 39, 1227–1236

(1994)

5. Geradts, Z., Keijzer, J.: The image-database rebezo for shoeprints

with developments on automatic classiﬁcation of shoe outsole

designs. Forensic Sci. Int. 82, 21–31 (1996)

6. Bouridane, A., Alexander, A., Nibouche, M., Crookes, D.: Appli-

cation of fractals to the detection and classiﬁcation of shoe-

prints. In: Proceedings International Conference on Image

Processing. Volume 1. pp. 474–477 (2000)

7. Alexander, A., Bouridane, A., Crookes, D.: Automatic classiﬁca-

tion and recognition of shoeprints. In: Proceedings Seventh

International Conference on (Conf Image Processing and Its

Applications Publ. No. 465). Volume 2. 638–641 (1999)

8. de Chazal, P., de Chazal, P., Flynn, J., Reilly, R.: Automated

processing of shoeprint images based on the fourier transform

for use in forensic science. Trans. Pattern Anal. Mach. Intell. 27,

pp. 341–350 (2005)

9. Su, H., Crookes, D., Bouridane, A., Gueham, M.: Local image

features for shoeprint image retrieval. In: British Machine Vision

Conference 2007. (2007)

10. Zhang, L., Allinson, N.: Automatic shoeprint retrieval system for

use in forensic investigations. In: 5th Annual UK Workshop on

Computational Intelligence (2005)

11. Everingham, M., Zisserman, A., Williams, C.K.I., Van Gool, L.:

The PASCAL Visual Object Classes Challenge 2006 ( VOC2006)

Results (2006)

12. Pavlou, M., Allinson, N.: Automatic extraction and classiﬁcation

of footwear patterns. In: Intelligent Data Engineering and Auto-

mated Learning, IDEAL 2006. pp. 721–728 (2006)

13. Su, H., Crookes, D., Bouridane, A.: Thresholding of noisy shoe-

print images based on pixel context. Pattern Recognit. Lett. 28,

301–307 (2007)

14. Su, H., Bouridane, A., Crookes, D.: Image quality measures for

hierarchical decomposition of a shoeprint image. Forensic Sci.

Int. 163, 125–131 (2006)

Force Field Feature Extraction

The overall objective in deﬁning feature space is to

reduce the dimensionality of the original pattern

space, while maintaining discriminatory power for

classiﬁcation. To meet this objective in the context of

ear biometrics a novel force ﬁeld transformation

which treats the image as an array of mutually attract-

ing particles that act as the source of a Gaussian force

ﬁeld has been developed. Underlying the force ﬁeld

there is a scalar potential energy ﬁeld, which in the

case of an ear takes the form of a smooth surface that

562

Force Field Feature Extraction

resembles a small mountain with a number of peaks

joined by ridges. The peaks correspond to potential

energy wells and to extend the analogy the ridges

correspond to potential energy channels. Since the

transform also happens to be invertible, and since the

surface is otherwise smooth, information theory sug-

gests that much of the information is transferred to

these features, thus conﬁrming their efﬁcacy. Force

ﬁeld feature extraction, using an algorithm similar to

gradient descent, exploits the directional properties of

the force ﬁeld to automat ically locate these channels

and wells, which then forms the basis of the character-

istic ear features.

▶ Physical Analogies for Ear Recognition

Force Field Transform

An invertible linear transform which transforms an

image into a force ﬁeld by pretending that pixels have

a mutual attraction proportional to their intensities

and inversely to the square of the distance betw een

them rather like Newton’s Law of Universal Gravita-

tion. Each pixel is assumed to generate a spherically

symmetrical force ﬁeld so that the total force Fðr

exerted on a pixel of uni t intensity at the pixel location

with position vector r

by a remote pixel with position

vector r

and pixel intensities Pðr

Þ is given by the

vector summation,

Fðr

Þ¼

Pðr

 r



8i 6¼ j

08i ¼ j

;

: ð1Þ

To calculate the force ﬁeld for the entire image, this

equation should be applied at every pixel position in

the image. In practice this computation would be done

in the frequency domain using Eq. 2 where ℑ stands for

FFT and ℑ

1

stands for inverse FFT.

forcefield ¼

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

M N

ℑ

1

ℑ unitforcefieldðÞℑ imageðÞ½:

ð2Þ

▶ Physical Analogies for Ear Recognition

Forensic

Forensic is the use of science or technology in the

investigation and establishment of facts or evidence

in the court of l aw.

▶ Skull, Forensic Evidence of

Forensic Anthropology

Forensic anthropology is the application of physi-

cal anthropology in special cases with forensic impor-

tance, such as to identify the skeletonized human

remains.

▶ Skull, Forensic Evidence of

Forensic Applications, Overview

CHRISTOPHE CHAMPOD

Institut de Police Scientiﬁque, Ecole des Sciences

Criminelles, Universite

de Lausanne, Switzerland

Introduction

The use of biometric data is a decisive process in

▶ forensic science that helps to establish a person’s

identity or associate two unknown persons. Forensic

scientists realized that physiological or behavioral data

could help to inform about, sort, and potentially indi-

vidualize the persons involved in criminal offences. It

is the case when (1) an unknown individual (living or

his/her remains) has to be identiﬁed, (2) when biomet-

ric traces left by unknown individuals during activities

of interest have to be traced back to their sources, or

(3) when biometric traces have to be linked together in

a series. Situations (1) and (2) require comparison

between biometric information gathered from un-

known sources and material of known (or declared as

Forensic Applications, Overview

563

such) origin, either on a one-to-one or on a one-to-

many basis. In the latter case, data of known origin are

organized in a database, allowing one-to-many

searches. The third situation (3) compares biometric

material from unknown sources and groups them

according to potential (yet unidentiﬁed) sources. This

last activity may involve the use of a database or may be

carried out on a case-by-case basis.

Both situations (2) and (3) take advantage of what

is known in forensic science as ‘‘Locard’s exchange

principle’’. Locard suggested in 1920 that forensic

scientists can take advantage of the traces (such as

ﬁbers, paint, ﬁrearms discharge residues, dust, blood-

stains, etc.) and marks (e.g., ﬁnger marks, footwear

marks, toolmarks, etc.) exchanged between actors

(e.g., a victim and an offender), and associated objects

or scenes involved in criminal activities. The systematic

search (potentially helped with detection techniques),

preservation and analysis of these marks and traces will

help to establish the relation between the actors,

objects or scenes, and reconstruct the course of activ-

ities. The biometric features that can be helpful are

many (and tend to increase in our modern society)

and may consist of the following types of traces or

marks:

1. Handwritten notes, including disputed signatures.

2. Finger marks and, by extension, any marks of fric-

tion ridge skin.

3. Barefoot impressions.

4. Bloodstain, semen stain, saliva stain, and any other

biological ﬂuids.

5. Earmarks.

6. Impression left by a face on airbags in car crashes

[1].

7. Images of individuals (including images showi ng

face, ears, or other identifying features) in stills or

videos taken either from CCTV systems or any

image-recording device.

8. Recording of a voice utterance of an individual

(either in analog or digital form).

Biometric features have been used in forensic science

for many centuries for some attributes (such as hand-

writing) and more recently for others (e.g., DNA

proﬁling or recognition of faces from CCTV surveil-

lance camera). However, in many respects, the applica-

tion of biometry in forensic science is in contrast to the

deployment of biometric systems in other areas (such

as access control) this is mainly due to the

unpredictable nature in terms of quality and quantity

of the biometric features available to the forensic sci-

entist, especially when they are collected as traces.

These contrasts have been detailed elsewhere [2].

This overview is intended to cover the standard

use of biometric features by forensic experts, either

manually or with semi-au tomated or automated sys-

tems. While distinguishing the types of biometry

helps to structure the text, it also helps to distinguish

between the investigative and evaluative use of the

techniques.

In the investigative mode, marks and traces are

questioned to provide leads that may help to focus

the inquiry, without making any reference to a poten-

tial source at the outset. Typical questions are as

follows:

1. Can we link these criminal incidents the biometric

evidence recovered?

2. On the basis of the recovered material, can we make

any inference regarding the sex, age, ethnicity, or

other physical attributes of the donor?

3. By comparing this biometric entry with a forensic

database, is it possible to prepare a short list of

potential donors?

In the evaluative mode, the questions are directed

towards a source that is available, to provide control

material to be compared against the unknown materi-

al. The case is then focused on one or more identiﬁed

persons, with a view to help assess whether they are the

source of the recovered material.

The references used hereinafter are limited to

major textbooks or papers for each evidence type con-

sidered. Some evidence types are largely covered

in speciﬁc chapters of this encyclopedia and address

speciﬁcally the issues associated with automatic

recognition.

Admissibility of Biometric Evidence

in Court

The question of admissibility is crucial when the scien-

tiﬁc element is used in the evaluative mode, especially

when it is expected to ﬁnd and present identiﬁcation

evidence in a courtroom. In the United States of

America, the criterion for admissibility was tradition-

ally based on the Frye rule (1923), which invites the

judge – acting as a gatekeeper – to assess if the

564

Forensic Applications, Overview

technique has gained general acceptance within the

relevant scientiﬁc community. It was under Frye that

numerous human identiﬁcation evidence types such as

ﬁngerprints, handwriting evidence, and DNA gained

acceptance. Earprint, however, failed to pass the Frye

test. The Frye standard has been revised following a

ruling of the US Supreme court in Daubert v. Merrell

Dow Pharmaceuticals (1993) and its progeny [3].

Daubert gave an interpretation of the Federal Rule of

Evidence (FRE) and required the judge, still acting as

gatekeeper, to assess more than only the general accep-

tance and include ﬁve criteria:

1. Whether the expert’s technique or theory can be or

has been tested, that is, whether the expert’s theory

can be challenged in some objective sense.

2. Whether the technique or theory has been subject

to peer review publication.

3. The known or potential rate of error of the tech-

nique or theory when applied.

4. The existence and maintenance of standards and

controls.

5. Whether the technique or theory has been generally

accepted in the scientiﬁc community.

Nowadays in the United States of America, at the

federal level, Daubert is in force. States may apply

either Frye (or similar state decisions) or Daubert.

Daubert led to an increased number of challenges

in court and forced the forensic community to articu-

late in detail the foundations of their disciplines, even

in areas that had gained acceptance in US courtrooms

for numerous years (see http://www.daubertonthe

web.com/).

In Europe, there is no speciﬁc admissibility rule

regarding scientiﬁc evidence. The principle of the judges’

free evaluation of the evidence prevails. Hence, it is not

surprising to see currently, a limited debate in the

European jurisprudence regarding the admissibility of

identiﬁcation evidence.

From Anthropometry to Fingerprinting

and AFIS

In the face of the absence of any reliable means of

identifying recidivists, Alphonse Bertillon proposed

in 1881 a classiﬁcation and retrieval method based on

anthropological measures. He took advantage of the

fact that bone lengths remain constant in adulthood.

It varies from individual to individual and can be

measured with reasonable precision. Eleven precise

measurements (height of the individual, length of out-

stretched arms, height of trunk, length and width of

the head, length of the left middle ﬁnger, the left foot,

left forearm, and right ear) combined with a mention

of the color of the iris, were proposed to establish an

anthropometric form for each arrested individual. This

anthropometric record was completed with a photo-

graph of the face and a standardized description of

particular marks that can be ﬁlled and retrieved. The

system was essentially used as an investigative measure

to help identify individuals arrested multiple times

(in time and place). The combination of anthropomet-

ric measurement, forensic photograph, and the

standard description of the face was coined ‘‘Bertillon-

nage’’. A rapid spread of Bertillonnag e has been ob-

served at the turn of the 20th century across the world

police departments and penal institutions [4]. The

limitations of this techn ique were quickly noticed: (1)

uneven distributions of the measurements; (2) correla-

tion between features; (3) inter-operator varia tions,

and (4) the imperative need of the body of the individ-

ual because no anthropometric traces are left on crime

scenes (at the time).

In 1880, Herschel, a colonial administrator in India,

published his proposal to use ﬁnger prints to identify

individuals. At the same time (in 1880), Fau lds, a med-

ical missionary in Japan, proposed using ﬁnger prints

for investigative identiﬁcation purposes as well, as

ﬁnger marks could be detected on crime scenes. Finger-

printing became a credible alternative to anthropome-

try for identiﬁcation of habitual offenders, when

Galton presented in 1892 the basic axioms of ﬁnger-

printing, including permanence (based on Herschel’s

work and data), discriminative power (Galton pub-

lished the ﬁrst statistical model on the ﬁngerprint varia-

bility), and the possibility of reliably classifying

ﬁngerprints into three basic patterns. The classiﬁcation

method was then greatly improved by Henry and

gained a large acceptance in English-speaking

countries. Almost simultaneously, Vucetich , proposed

a simpler system, based on Galton’s initial proposal,

that proved very successful for small to medium size

databases. From the early 1900s, ﬁngerprinting became

the sole means of identiﬁcation of habitual offen ders

throughout the world [5]. The use here is investigative

(search for a potential candidate in a repository of

ﬁngerprint forms) and evaluative (veriﬁcation on a

Forensic Applications, Overview

565

one-to-one basis), based on ten print records (taken

from living or dead individuals).

An additional beneﬁt of papillary ridge skin is that

marks are left on the crime scene and are either readily

visible or detectable using adequate dete ction techni-

ques [6]. The ﬁrst cases of identiﬁcation of criminals

through the ﬁnger marks left by them are attributed to

Vucetich (1892) and Bertillon (1902). These marks can

be searched against a ﬁngerprint ﬁle or, of late, in an

AFIS (Automatic Fingerprint Identiﬁcation System).

With the increase of tenprint cards and the difﬁculty

of searches based on papillary marks, research in auto-

matic retrieval processing systems took off in parallel

with the technological advances since the 1960s. All

forensic AFIS are nowadays largely based on minutiae

matching both for ﬁnger and palm impressions [7].

The main advantage of an AFIS is the ability to com-

pare a single print or mark, as well as a tenprint card, to

the whole database. Note that AFIS provides a list of

best candidates (according to a scoring/ranking met-

ric). The identiﬁcation process is not carried out by the

system, but processed manually by an expert (through

a dedicated user interface) in exactly the same way as if

the potential candidate prints were sug gested as a

result of the usual police inquiry.

Other Biometric Characteristics used for

Human Identification Purposes

The use of deoxyribonucleic acid (DNA), a chain of

nucleotides contained in the nucleus of our cells, has

been a major breakthrough in forensic science to help

in the identiﬁcation of unknown individuals or

biological samples left by them. Nuclear DNA can be

extracted from all biological tissues (blood, saliva,

urine or semen, from hair (with roots) and skin cells

left by contact with the skin). For identiﬁcation pur-

poses (of the liv ing or dead), samples are obtained

from blood, saliva, or bones. The most common anal-

ysis of nuclear DNA is focused on STRs (short tandem

repeats) [8]. STRs are repetitive sequences at a given

location of the DNA molecule, of non-coding nature,

which show a large and well-documented polymor-

phism. At a given locus, one individual will show two

speciﬁc numbers of repetitions of the given sequence of

nucleotides. These two numbers called alleles then give

the biometric template for that locus. Note that one

allele results from the genetic transmission from

the biological father, and the other from the biolog-

ical mother. The constitution of DNA databases to

assist investigation is simple and has been developed

in most countries.

Forensic applications of DNA proﬁling for human

identiﬁcation are numerous for STR analysis and cover

(1) the comparison of an unknown proﬁle (for exam-

ple, from human remains) of an individual to a data-

base of known proﬁles or proﬁles of potential relatives;

(2) ﬁliation testing when putative genitors are available

or alternatively w ith ancestors or descendants. The

ﬁrst introduction of DNA proﬁling in forensic science

dates back to 1986 (the Pitchfork case in the UK),

but the large development of practices started in

the 1990s. Before that time, biological ﬂuids were ana-

lyzed using blood grouping determination or analysis

of various proteins or enzymes [9]. Most of these

forensic analyses have been abandoned for identiﬁca-

tion purposes in favor of DNA proﬁling because

of the limited sensitivity and discriminating power

of these systems.

When nuclear DNA cannot be analyzed (typically

because of the degradation of DNA), mitochondrial

DNA (contained in the mitochondria and inherited

through maternal lines) can be used. Its discriminat-

ing power, however, is much lower than STR nuclear

DNA analysis.

Dental features are mainly used in the identi ﬁcation

of human remains in cases of missing persons or mass

disasters [10]. The features used range from the stan-

dard dental record (indication of missing teeth, resto-

rations, crowns, etc.) to dental radiographs (tooth

contours, relative positions of neighboring teeth, and

shapes of the dental work). These anatomical features

have shown very good stability and variability and the

teeth serve as a suitable repository of manmade opera-

tions that will leave various marks and shapes. Alpha-

numerical data can easily be organized in databases

and such systems are used operationally in cases

of mass disasters (http://www.interpol.in t/Public/

DisasterVictim/Default.asp).

Forensic analysis of soft tissues can help in the iden-

tiﬁcation of remains. Analysis of scars, incised wounds,

burn marks, trauma, and medical/surgical intervention

are typical either in the investigative or evaluative

mode [11]. When no tissue is left, forensic anthropology

becomes an essential part of forensic and archeological

investigations [11, 12]. Following the recovery of un-

identiﬁed skeletal remains, the forensic anthropologist

566

Forensic Applications, Overview

can assist in guiding the investigation to identify the

sex, ethnic origin, stature, and age (and if it is a

woman’s remains, whether she had gave birth) of the

deceased. This information is investigative in nature.

The same applies to cranio-facial reconstruction from

the skull to help in the search for a deceased person

[13]. When reference material is made available (X-ray

images from ante mortem medical documentation), its

comparison w ith post mortem data can help to estab-

lish identity through the analysis of morphology, frac-

tures, medical interventions on the bones, and frontal

sinus shapes [14]. Most of these areas have not been

subjected to extensive automation research [15].

Other Biometric Marks Left Following

Activities of Forensic Interest

DNA proﬁles can be obtained from the marks left

behind by activities of forensic interest, typically from

stains of blood, saliva, urine or semen, and from hair

(with roots) and skin cells left by mere contact.

Extracts are ampliﬁed using a sensitive and selective

DNA replication method known as Polymerase Chain

Reaction (PCR). In practice sensitivity to levels below

100 pg of DNA (a few cells) can be achieved. Such

sensitivity widens the investigative possibilities, allow-

ing the analysis of biological stains of very limited

quantity. These proﬁles can be used to compare a

DNA proﬁle obtained from biological material against

proﬁles from known individuals. If a correspondence is

obtained, then this information can be used as evi-

dence in court. It is important to stress that a match

between two DNA proﬁles does not establish conclu-

sively an identiﬁcation of sources. Indeed, although the

selectivity of DNA proﬁling is very high, there exists a

probability of random association. In addition, DNA

analysis offers some investigative capabilities gathered

through the systematic comparison of DNA proﬁles

coming from various scenes or familial searches against

the DNA database. Another investigative aspect is the

use of speciﬁc DNA analysis (SNP for single nucleotide

polymorph ism) to infer (w ithin deﬁned uncertainty

boundaries) iris or hair color, skin pigmentation, and

ancestry background [16].

The morphology of the ear was considered by

Bertillon as the most identifying part of an individual.

This modality was thus quickly used for identiﬁcation

purposes in forensic cases, either on photographs

(or still images from video recording [17]) or on ear

marks left on crime scenes, for instance, on doors.

Forensic ear or ear print comparison is traditionally

completed by skilled examiners according to published

principles and protocols [18]. It is important to note

that there is a big gap in terms of quality between a

well-taken pho togra ph of a ear and its i mpression on

a door; hence, the strength of the evidence may vary

from case to case as a function of the quality and the

extent of the available material. Ear print examination

can be used in the investigative phase to constitute a

series based on the collected marks or to estimate the

height of the donor, or in an evaluative manner, to

associate a recovered mark with the ears of a designated

individual.

Barefoot impressions can be left either on crime

scenes or inside the shoes [19]. In the ﬁrst instance,

their investigation will help to assess the sequence of

events and associate or exclude a given individual from

being at the source of these marks. In the second, the

analysis can help to assess whether a given individual is

the habitual wearer of the shoe. Their use in criminal

investigations predates the use of ﬁnger marks [20].

Barefoot impressions have shown a very high discrim-

ination power and allow, when the quality of the mark

is adequate, to bring powerful evidence of the identity

of the sources in court.

Bite marks can be left on various substrates (the

skin of a victim, some food, etc.) and can be compared

against the control material from potential donors.

A full account of their detection and analysis can be

found in [21]. If bite mark analysis is to continue to

play a role in the judicial process, there is an urgent

need for high quality studies that meet the levels of

forensic and scientiﬁc scrutiny applied to the other

disciplines w ithin the criminal justice system [22, 23].

From time to time, lip marks can be recovered from

objects that came into contact with lips. Forensic lip

print analysis is a very anecdotal area.

Handwriting and signature are biometric attributes

with a long history in forensic science. The principles

and procedures used by forensic experts to assign ques-

tionable handwritten documents to known individuals

are described in [24]. The forensic expert tries to assess

existing similarities and dissimilari ties between control

and recovered samples through a subjective estimation

of the individuality and variability of the material at

hand. At the moment, the automatic techniques used

for handwriting and signature recognition are in their

Forensic Applications, Overview

567

infancy, especially in forensic science. Following the

Daubert challenges, the ﬁeld has been the focus of an

increased scrutiny as to its scientiﬁc underpinning. It

has led to a new body of research that shows the fertile

avenues of collaborati on between biometric computer

science and forensic science [25–27].

Analog/Digital Biometric Information

Recorded in Investigations

Forensic speaker recognition can be deﬁned as any

process using speech signals to determine if a speciﬁc

individual was the speaker of a speciﬁc declaration.

Experts may reach opinions from a variety of tech-

niques used alone or in combination: auditive com-

parison, visual comparison of spectrograms, and

semi-automatic methods for extraction of speciﬁc

parameters (e.g., formant frequencies). Auditive

comparisons are more likely to be conducted by pho-

neticians. They assess voice characteristics (voice,

speech, language, and linguisti c) either subjectively

or objectively (using signal processing tools). The

visual spectrographic approach was ﬁrst proposed in

1962. In 1976, the US National Academy of Sciences

recommended the use of this approach in forensic

cases cautiously [28]. Automatic speaker recognition is

also used in for ensic science (see related entry in this

Encyclopedia). Several characterization and modeling

tools have been developed for automatic speaker recogni-

tion. All are sensitive to voice modiﬁcation in recording

and transmission conditions and their performance wor-

sens when the conditions deteriorate. In forensic cases,

the rec ording conditions of the trace and the reference

materials are rarely similar or ideal, but rather record in

different and unconstrained conditions, i.e., thr o ugh mo-

bile communications (GSM) transmission and with

background noise. Due to these factors, the comparison

is often undertaken under adverse conditions.

Facial images are more and more available for

forensic investigations. Forensic face recognition is

generally carried out by dedicated experts using appro-

aches based either on morphological analysis of facial

structures, anthropometric meas urements, or image

superimposition [29]. The morphological approach

is based on a nomenclature for the description of

the physiological aspe cts of the nose, the forehead,

and the ear. Additional information, such as facial

wrinkles and scars, can also be used. As the description

is rather subjective, variations between operators are

observed. In addition, the features of the same individ-

ual change due to expression changes, photographic

angles or aging, and the demonstration of their statis-

tical independence is often weak. The anthropometric

approach can be described as the quantiﬁcation of

physiological proportions between speciﬁc facial land-

marks. This method is only used for the comparison of

faces with the same orientation. In order to avoid any

scale and absolute size differences between photographs,

ratios are calculated from these landmarks. Lighting

conditions, camera distortions, camera positioning,

facial orientation, facial expressions, and aging may

impact the measures. The superimposition-based

approach is the juxtaposition or the superimposition

of facial images, taken under the same acquisition con-

ditions (the orientation, the pose, and the size).

These three main comparison approaches do not

yet consider automatic face recognition (a subject cov-

ered in several chapters of this Encyclopedia). Auto-

matic face recognition systems have a large role to play

in the future, not only in dealing with the face as such

but also taking advantage of lips [30] or other features.

But before introducing any automatic face recognition

in court, a full and systematic assessment of the system

should be conducted under realistic conditions, using

ﬁt-for-purpose forensic efﬁciency measures.

The prevalence of images or videos in modern

society opens the route to the development of new

types of forensic biometry (some are already covered

in this Encyclopedia, e.g. gait analysis). Is it not rare to

observe anatomical features on images that can help

towards the identiﬁcation of the individual captured

on these images? Thes e features can be skin details,

scars, veins and tattoos. Biometric developments are

still in their early stages [31].

References

1. Yamazaki, K., Imaizumi, K., Kubota, S., Atsuchi, M., Noguchi, K.,

Yosino, M.: Experimental study on personal identiﬁcation from

faceprint on vehicle’s airbag. Japanese Journal of Science and

Technology of Identiﬁcation. 9(1), 19–27 (2004)

2. Dessimoz, D., Champod C.: Linkages between biometrics and

forensic science. In: Flynn, P.J, Jain, A.K, Ross, A. (eds.) Hand-

book of Biometrics, pp. 425–459. Springer, New York (2007)

3. Saks, M.J., Faigman, D.L.: Expert evidence after Daubert. Annu.

Rev. Law Soc. Sci. 1(1), 105–130 (2005)

568

Forensic Applications, Overview

4. Cole, S.: Suspect identities: A history of ﬁngerprinting and

criminal identiﬁcation. Harvard University Press, Cambridge,

MA (2001)

5. Berry, J., Stoney, D.A.: The history and development of ﬁnger-

printing. In: Lee, H.C., Gaensslen, R.E. (eds.) Advances in Fin-

gerprint Technology, 2nd edn. pp. 1–40. CRC Press, Boca Raton,

FL (2001)

6. Champod, C., Lennard, C.J., Margot, P.A., Stoilovic, M.: Finger-

prints and other Ridge Skin Impressions. CRC Press, Boca

Raton, FL (2004)

7. Komarinski, P.: Automated ﬁngerprint identiﬁcation systems

(AFIS). Elsevier, New York (2005)

8. Butler, J.M.: Forensic DNA typing, 2nd edn. Elsevier, Burlington,

MA (2005)

9. Gaensslen, R.E.: Sourcebook in Forensic Serology, Immunology,

and Biochemistry. US Department of Justice, National Institute

of Justice, US Printing Ofﬁce, Washington, DC (1983)

10. Sweet, D., Pretty, I.A.: A look at forensic dentistry – Part 1: The

role of teeth in the determination of human identity. Br.

Dent. J. 190(7), 359–366 (2001)

11. Black, S.M. (ed.): Forensic Human Identiﬁcation: An Introduc-

tion. CRC Press, Boca Raton, FL (2006)

12. Pickering, R.B., Bachman, D.C.: The Use of Forensic Anthropol-

ogy. CRC Press, Boca Raton, FL (2000)

13. Iscan, M.Y., Helmer, R.P. (ed.): Forensic Analysis of the Skull:

Craniofacial Analysis, Reconstruction, and Identiﬁcation. Wiley,

New York (1993)

14. Christensen, A.M.: Assessing the variation in individual

frontal sinus outlines. Am. J. Phys. Anthropol. 127(3), 291–295

(2005)

15. Falguera, J.R., Falguera, F.P.S., Marana, A.N.: Frontal sinus rec-

ognition for human identiﬁcation. In: Vijaya Kumar, B.V.K.,

Prabhakar, S., Ross, A.A. (eds.) Biometric Technology for

Human Identiﬁcation V. In: Proceedings of the SPIE; 2008

March 18, 2008; Orlando, FL. SPIE; 2008. p. 69440S–9

16. Frudakis, T.: Molecular photoﬁtting: Predicting Ancestry

and Phenotype using DNA. Academic Press, Burlington, MA

(2008)

17. Hoogstrate, A.J., van den Heuvel, C., Huyben, E.: Ear identiﬁca-

tion based on surveillance camera images. Sci. Justice. 41(3),

167–172 (2001)

18. van der Lugt, C.: Earprint Identiﬁcation. Elsevier Bedrijfsinfor-

matie, Gravenhage (2001)

19. Kennedy, R.B., Yamashita, A.B.: Barefoot morphology compari-

son: A summary. J. Forensic Ident. 57(3), 383–413 (2007)

20. Causse

, S.: Des empreintes sanglantes des pieds, et de leur mode

de mensuration. Annales d’hygie

ne publique et de me

decine

gale. 1854;1 (2e

me se

rie):175–89

21. Dorion, B.J. (ed.): Bitemark Evidence. Marcel Dekker, New York

(2005)

22. Pretty, I.A.: The barriers to achieving an evidence base for

bitemark analysis. Forensic Sci. Int. 159(Suppl 1), S110–S20

(2006)

23. Bowers, C.M.: Problem-based analysis of bitemark misidentiﬁ-

cations: The role of DNA. Forensic Sci. Int. 159(Suppl 1),

S104–S9 (2006)

24. Huber, R.A., Headrick, A.M.: Handwriting Identiﬁcation: Facts

and Fundamentals. CRC Press, Boca Raton, FL (1999)

25. Marquis, R., Schmittbuhl, M., Bozza, S., Taroni, F.: Quantitative

characterization of morphological polymorphism of handwrit-

ten characters loops. Forensic Sci. Int. 164, 211–220 (2006)

26. Schomaker, L.: Advances in writer identiﬁcation and veriﬁca-

tion. In: Ninth International Conference on Document Analysis

and Recognition – ICDAR 2007, pp. 1268–1273 (2007)

27. Srihari, S., Huang, C., Srinivasan, H.: On the discriminability of

the handwriting of twins. J. Forensic Sci. 53(2), 430–446 (2008)

28. Bolt , R.H., Coo per, F.S., G reen, D.M., Hamlet, S.L.,

McKnight, J.G., Pickett, J.M. et al.: On the Theory and Practice

of Voice Identiﬁcation. National Research Council, National

Academy of Sciences, Washington, DC (1979)

29. Iscan, M.Y.: Introduction of techniques for photographic com-

parison: Potential and problems. In: Iscan, M.Y., Helmer, R.P.

(eds.) Forensic Analysis of the Skull, pp. 57–70. Wiley-Liss, Inc.,

New York (1993)

30. Choras

, M.: Human lips as emerging biometrics modality. In:

Image Analysis and Recognition: 5th International Conference,

ICIAR 2008, Po

voa de Varzim, Portugal, June 25–27, 2008 Pro-

ceedings, p. 993–1002. Springer, Berlin (2008)

31. Jain, A., Lee, J-E., Jin, R.: Tattoo-ID: Automatic tattoo image

retrieval for suspect and victim identiﬁcation. In: Advances in

Multimedia Information Processing – PCM 2007, pp. 256–265

(2007)

Forensic Barefoot Comparisons

BRIAN A. YAMASHITA

,ROBERT B. K ENNEDY

Forensic Identiﬁcation Operations Support Services,

National Services and Research, Royal Canadian

Mounted Police, Ottawa, ON, Canada

Royal Canadian Mounted Police (retired), Ottawa,

ON, Canada

Synonyms

Barefoot morphology comparison; Footprint

comparison

Definitions

Forensic barefoot comparison, or barefoot morpholo-

gy comparison, describes the comparison of impres-

sions of the weight-bearing areas of feet in an attempt

to include or exclude a suspect as someone linked to a

crime scene. A bare or socked foot impression found at

the crime scene can be compared to inked barefoot

Forensic Barefoot Comparisons

569