You are on page 1of 20

Presented by:-

RAMSWAROOP SINGH T PRAMOD KUMAR REDDY P


BRANCH: CSE BRANCH: CSE
ROLL NO: 05C71A0547 ROLL NO: 05C71A0514
CONTACT NO: 9966952101 CONTACT NO: 9866562166
EMAIL ID: ram_fr_547@yahoo.co.in
ELLENKI COLLEGE OF ENGG. & TECH., PATEL GUDA
Abstract
“Biometrics” is a general term used Registered Traveler (RT) program.
alternatively to describe a Many companies are also
characteristic or a process. implementing biometric technologies
As a characteristic: to secure areas, maintain time records,
1. A measurable biological and enhance user convenience. For
(anatomical and physiological) and example, for many years Disney World
behavioral has employed biometric devices for
characteristic that can be used season ticket holders to expedite and
for automated recognition. simplify the process of entering its
As a process: parks, while ensuring that the ticket is
2. Automated methods of used only by the individual to whom it
recognizing an individual based on was issued.
measurable A typical biometric system is
biological (anatomical and comprised of five integrated
physiological) and behavioral components: A sensor is used to
characteristics. collect the data and convert the
Biometric systems have been information to a digital format. Signal
researched and tested for a few processing algorithms perform
decades, but have only recently entered quality control activities and develop
into the public consciousness because the biometric template. A data
of high profile applications, usage in storage component keeps information
entertainment media (though often not that new biometric templates will be
realistically) and increased usage by compared to. A matching algorithm
the public in day-to-day activities. compares the new biometric template
Example deployments within the to one or more templates kept in data
United States Government include the storage. Finally, a decision process
FBI’s Integrated Automated (either automated or human-assisted)
Fingerprint Identification System uses the results from the matching
(IAFIS), the US-VISIT program, the component to make a system-level
Transportation Workers Identification decision.
Credentials (TWIC) program, and the

Biometric Modalities
Commonly implemented or studied implementing a biometric device
biometric modalities include including location, security risks, task
fingerprint, face, iris, voice, signature (identification or verification),
and hand geometry. Many other expected number of users, user
modalities are in various stages of circumstances, existing data, etc. It is
development and assessment. There is also important to note that biometric
not one biometric modality that is best modalities are in varying stages of
for all implementations. Many factors maturity.
must be taken into account when

Dynamic Signature
Introduction signature and are common in locations
“Dynamic Signature” is a where merchants are capturing
biometric modality that uses, for signatures for transaction
recognition purposes, the anatomic and authorizations. Data such as the
behavioral characteristics that an dynamically captured direction, stroke,
individual exhibits when signing his or pressure, and shape of an individual’s
her name. Dynamic devices should not signature can enable handwriting to be
confuse with the electronic signature a reliable indicator of an individual’s
capture systems that are used to identity
capture a graphic image of the

Figure 1: Dynamic Signature


Depiction: As an
Individual signs the contact sensitive
tablet
Various measurements are observed and processed for comparison

Approach these characteristics in their analyses.


Dynamic signature Common dynamic characteristics
recognition uses multiple include the velocity, acceleration,
characteristics in the analysis of an timing, pressure, and direction of the
individual’s handwriting. These signature strokes, all analyzed in the X,
characteristics vary in use and Y, and Z directions.
importance from vendor to vendor and Figure 2 illustrates these
are collected using contact sensitive recorded Dynamic characteristics of a
technologies, such as PDAs or signature. The X and Y position are
digitizing Tablets used to show the changes in velocity in
the respective directions (indicated by
Most of the features used are
the white and yellow lines) while the Z
dynamic characteristics rather than
direction (red line) is used to indicate
static and geometric characteristics,
changes in pressure with respect to
although some vendors also include
time.

Figure 2: Graphic Depiction of Dynamic Signature Characteristic


Some dynamic signature recognition to the handwriting style of the
algorithms incorporate a learning individual. Despite this major strength
function to account for the natural of dynamic signature recognition, the
changes or drifts that occur in an characteristics historically have a large
individual’s signature over time.1The intra-class variability (meaning that an
characteristics used for dynamic individual’s own signature may vary
signature recognition are almost from collection to collection), often
impossible to replicate. Unlike a making dynamic signature recognition
graphical image of the signature, difficult. Recent research has reported
which can be replicated by a trained that static writing samples can be
human forger, a computer successfully analyzed to overcome this
manipulation, or a photocopy, dynamic issue.
characteristics are complex and unique

Face Recognition
Introduction verification and identification

Humans often use faces to recognize Predominant Approaches


individuals and advancements in As researcher interest in face
computing capability over the past few recognition continued, many different
decades now enable similar algorithms were developed, three of
recognitions automatically. Early face which have been well studied in face
recognition algorithms used simple recognition literature: Principal
geometric models, but the recognition Components Analysis (PCA), Linear
process has now matured into a science Discriminant Analysis (LDA), and
of sophisticated mathematical Elastic Bunch Graph Matching
representations and matching (EBGM).
processes. Major advancements and PCA: Principal Components
initiatives in the past ten to fifteen Analysis (PCA)
years have propelled face recognition PCA commonly referred to as the use of
technology into the spotlight. Face eigenfaces, is the technique pioneered
recognition can be used for both by Kirby and Sirivich in 1988. With
PCA, the probe and gallery images image may be represented as a
must be the same size and must first be weighted sum (feature vector) of the
normalized to line up the eyes and eigenfaces, which are stored in a 1D
mouth of the subjects within the array. A probe image is compared
images. The PCA approach is then used against a gallery image by measuring
to reduce the dimension of the data by the distance between their respective
means of data compression basics and feature vectors. The PCA approach
reveals the most effective low typically requires the full frontal face to
dimensional structure of facial patterns. be presented each time; otherwise the
This reduction in dimensions removes image results in poor performance. The
information that is not useful and primary advantage of this technique is
precisely decomposes the face structure that it can reduce the data needed to
into orthogonal (uncorrelated) identify the individual to 1/1000th of the
components known as data presented.
eigenfaces. Each face

Figure: Standard Eigenfaces: Feature vectors are derived using Eigenfaces

LDA: Linear Discriminant Analysis class (i.e., within user) variance. In


LDA is a statistical approach for Figure 2 where each block represents a
classifying samples of unknown class, there are large variances between
classes based on training samples with classes, but little variance within
known classes. This technique aims to classes. When dealing with high
maximize between-class (i.e., across dimensional face data, this technique
users) variance and minimize within- faces the small sample size problem
that arises where there are a small compared to the dimensionality of the
number of available training samples sample space.

Figure: Example of Six Classes Using LDA


EBGM: Elastic Bunch Graph given pixel. It is the result of a
Matching
EBGM relies on the concept that real convolution of the image with a Gabor

face images have many non-linear filter, which is used to detect shapes

characteristics that are not addressed and to extract features using image

by the linear analysis methods processing. [A convolution expresses

discussed earlier, such as variations in the amount of overlap from functions,

illumination (outdoor lighting vs. blending the functions together.]

indoor fluorescents), pose (standing Recognition is based on the similarity

straight vs. leaning over) and of the Gabor filter response at each

expression (smile vs. frown). A Gabor Gabor node. This biologically-based

wavelet transform creates a dynamic method using Gabor filters is a process

link architecture that projects the face executed in the visual cortex of higher

onto an elastic grid. The Gabor jet is a mammals. The difficulty with this

node on the elastic grid, notated by method is the requirement of accurate

circles on the image below, which landmark localization, which can

describes the image behavior around a sometimes be achieved by combining


PCA and LDA methods.
Figure: Elastic Bunch Map Graphing

Fingerprint Recognition
Introduction biometric) due to advancements in
Fingerprint identification is one of the computing capabilities. Fingerprint
most well-known and publicized identification is popular because of the
biometrics. Because of their inherent ease in acquisition, the
uniqueness and consistency over time, numerous sources (ten fingers)
fingerprints have been used for available for collection, and their
identification for over a century, more established use and collections by law
recently becoming automated (i.e. a enforcement and immigration.

Approach
Concept characteristics sometimes used during
A fingerprint usually appears as the automatic classification and
a series of dark lines that represent the minutiae extraction processes. The
high, peaking portion of the friction types of information that can be
ridge skin, while the valleys between collected from a fingerprint’s friction
these ridges appears as white space and ridge impression include the flow of
are the low, shallow portion of the the friction ridges (Level 1 Detail), the
friction ridge skin. Fingerprint presence or absence of features along
identification is based primarily on the the individual friction ridge paths and
minutiae, or the location and direction their sequence (Level 2 Detail), and
of the ridge endings and bifurcations the intricate detail of a single ridge
(splits) along a ridge path. The images (Level 3 Detail). Recognition is
below present examples of fingerprint usually based on the first and second
features: (a) two types of minutiae and levels of detail or just the latter.
(b) examples of other detailed AFIS technology exploits some of
these fingerprint features. Friction overall ridges to assign a fingerprint
ridges do not always flow continuously classification and then extract the
throughout a pattern and often result in minutiae detail – a subset of the total
specific characteristics such as ending amount of information available yet
ridges, dividing ridges and dots, or enough information to effectively
other information. An AFIS is search a large repository of
designed to interpret the flow of the fingerprints.

Figure: Minutiae Figure: Other Fingerprint Characteristics

Hardware has significantly less capacitance than


A variety of sensor types — optical, an area of finger (friction ridge skin).
capacitive, ultrasound, and thermal — Other fingerprint sensors capture
are used for collecting the digital images by employing high frequency
image of a fingerprint surface. Optical ultrasound or optical devices that use
sensors take an image of the prisms to detect the change in light
fingerprint, and are the most common reflectance related to the fingerprint.
sensor today. The capacitive sensor Thermal scanners require a swipe of a
determines each pixel value based on finger across a surface to measure the
the capacitance measured, made difference in temperature over time to
possible because an area of air (valley) create a digital image.

Hand Geometry
Introduction 1980s. The systems are widely
Hand geometry recognition is the implemented for their ease of use,
longest implemented biometric type, public acceptance, and integration
debuting in the market in the late capabilities. One of the shortcomings
of the hand geometry characteristic is systems use a camera to capture a
that it is not highly unique, limiting the silhouette image of the hand (Figure
applications of the hand geometry 2). The hand of the subject is placed on
system to verification tasks only. the plate, palm down, and guided by
five pegs that sense when the hand is
Approach in place. The resulting data capture by
The devices use a simple concept of a Charge-Coupled Device (CCD)
measuring and recording the length, camera of the top view of the hand
width, thickness, and surface area of an including example distance
individual’s hand while guided on a measurements
plate (Figure 1). Hand geometry

Figure 1: Bottom View Figure 2: Silhouette of Hand Image

The image captures both the top storage needs of other biometric
surface of the hand and a side image systems.
that is captured using an angled mirror
(Figure 3). Upon capture of the The enrollment process of a hand
silhouette image, 31,000 points are geometry system typically requires the
analyzed and 90 measurements are capture of three sequential images of
taken; the measurements range from the hand, which are evaluated and
the length of the fingers, to the measured to create a template of the
distance between knuckles, to the user’s characteristics. Upon the
height or thickness of the hand and submission of a claim, the system
fingers (Figure 4). This information is recalls the template associated with
stored in nine bytes of data, an that identity; the claimant places
extremely low number compared to the his/her hand on the plate; and the
system captures an image and creates a
verification template to compare to the on the threshold of the system; the
template developed upon enrollment. A claim is either accepted or rejected
similarity score is produced and, based
.

Figure 3: Hand Including Mirror Figure 4: Example Distance Measurements.


Image as Seen by the CCD Camera
The enrollment process of a hand that identity; the claimant places
geometry system typically requires the his/her hand on the plate; and the
capture of three sequential images of system captures an image and creates a
the hand, which are evaluated and verification template to compare to the
measured to create a template of the template developed upon enrollment. A
user’s characteristics. Upon the similarity score is produced and, based
submission of a claim, the system on the threshold of the system; the
recalls the template associated with claim is either accepted or rejected.

Iris Recognition
Introduction iris is a muscle within the eye that
Iris recognition is the process of regulates the size of the pupil,
recognizing a person by analyzing the controlling the amount of light that
random pattern of the iris (Figure 1). enters the eye. It is the colored portion
The automated method of iris of the eye with coloring based on the
recognition is relatively young, amount of melatonin pigment within
existing in patent only since 1994.The the muscle (Figure 2).
Figure 1: Iris Diagram. Figure 2: Iris Structure

Although the coloration Approach


and structure of the iris Before recognition of
is genetically linked, the iris takes place, the
the details of the iris is located using
patterns are not. The landmark features.
iris develops during prenatal growth These landmark features and the
through a process of tight forming and distinct shape of the iris allow for
folding of the tissue membrane. Prior imaging, feature isolation, and
to birth, degeneration occurs, resulting extraction. Localization of the iris is
in the pupil opening and the random, an important step in iris recognition
unique patterns of the iris. Although because, if done improperly, resultant
genetically identical, an individual’s noise (e.g., eyelashes, reflections,
irides are unique and structurally pupils, and eyelids) in the image may
distinct, which allows for it to be used lead to poor performance.
for recognition purposes.

Figure 3: White outlines indicate the localization


of the iris and eyelid boundaries.

Iris imaging requires use of a high the subject. Upon imaging an iris, a 2D
quality digital camera. Today’s Gabor wavelet filters and maps the
commercial iris cameras typically use segments of the iris into phasors
infrared light to illuminate the iris (vectors). These phasors include
without causing harm or discomfort to information on the orientation and
spatial frequency (“what” of the information is used to map the Iris
image) and the position of these areas Codes® (Figures 4 & 5).
(“where” of the image).This

Figure 5: Pictorial Representation of Iris Code.

Iris patterns are described by an Iris two Iris Codes. If the HD indicates that
Code using phase information less than one-third of the bytes in the
collected in the phasors. The phase is Iris Codes are different, the Iris Code
not affected by contrast, camera gain, fails the test of statistical significance,
or illumination levels. The phase indicating that the Iris Codes are from
characteristic of an iris can be the same iris. Therefore, the key
described using 256 bytes of data using concept to iris recognition is failure of
a polar coordinate system. Also the test of statistical independence.
included in the description of the iris
are control bytes that are used to Iris vs. Retina Recognition
exclude eyelashes, reflection(s), and As discussed above, iris
other unwanted data. To perform the recognition utilizes the iris muscle to
recognition, two Iris Codes are perform verification. Retinal
compared. The amount of difference recognition uses the unique pattern of
between two Iris Codes — Hamming blood vessels on an individual’s retina
Distance (HD) — is used as a test of at the back of the eye. The figure
statistical independence between the below illustrates the structure of the
eye
Figure 6: Structure of the Eye.
Both techniques involve photochemical effects, the principal
capturing a high quality picture of the potential damage modality is thermal.
iris or retina, using a digital camera. In When NIR is produced using light
the acquisition of these images, some emitting diodes, the resulting light is
form of illumination is necessary. Both incoherent. Any risk for eye safety is
techniques use NIR (near infrared) remote with a single LED source using
light. Although safe in a properly today's LED technology. Multiple
designed system, eye safety is a major LED illuminators can, however,
concern for all systems that illuminate produce eye damage if not carefully
the eye. Because infrared has designed and used.
insufficient energy to cause

Palm Print Recognition


Introduction of friction ridge impressions either
Palm print recognition originated from the same source or
inherently implements many of the could not have been made by the same
same matching characteristics that source. Because fingerprints and
have allowed fingerprint recognition to palms have both uniqueness and
be one of the most well-known and permanence, they have been used for
best publicized biometrics. Both palm over a century as a trusted form of
and finger biometrics are represented identification. However, palm
by the information presented in a recognition has been slower in
friction ridge impression. This becoming automated due to some
information combines ridge flow, ridge restraints in computing capabilities and
characteristics, and ridge structure of live-scan technologies. This paper
the raised portion of the epidermis. provides a brief overview of the
The data represented by these friction historical progress of and future
ridge impressions allows a implications for palm print biometric
determination that corresponding areas recognition.
Concept recognition concept, one must first
Palm identification, just like understand the physiology of the ridges
fingerprint identification, is based on and valleys of a fingerprint or palm.
the aggregate of information presented When recorded, a fingerprint or palm
in a friction ridge impression. This print appears as a series of dark lines
information includes the flow of the and represents the high, peaking
friction ridges (Level 1 Detail), the portion of the friction ridged skin
presence or absence of features along while the valley between these ridges
the individual friction ridge paths and appears as a white space and is the
their sequences (Level 2 Detail), and low, shallow portion of the friction
the intricate detail of a single ridge ridged skin. This is shown in Figure 1.
(Level 3 detail). To understand this

Figure 1: Fingerprint Ridges (Dark Lines) vs. Fingerprint Valleys (White Lines)
Palm recognition technology exploits effectively search a large repository of
some of these palm features. Friction palm prints. Minutiae are limited to
ridges do not always flow continuously the location, direction, and orientation
throughout a pattern and often result in of the ridge endings and bifurcations
specific characteristics such as ending (splits) along a ridge path. The images
ridges or dividing ridges and dots. A in Figure 2 present a pictorial
palm recognition system is designed to representation of the regions of the
interpret the flow of the overall ridges palm, two types of minutiae, and
to assign a classification and then examples of other detailed
extract the minutiae detail — a subset characteristics used during the
of the total amount of information automatic classification and minutiae
available, yet enough information to extraction processes
Figure 2: Palm Print and Close-up
Hardware slow to adapt to the larger capture
areas required for digitizing palm
A variety of sensor types — capacitive,
prints. Challenges for sensors
optical, ultrasound, and thermal — can
attempting to attain high-resolution
be used for collecting the digital image
palm images are still being dealt with
of a palm surface; however, traditional
today.
live-scan methodologies have been

Speaker Recognition
Introduction shape of an individual’s vocal tract,
Speaker, or voice, recognition is a which consists of an airway and the
biometric modality that uses an soft tissue cavities from which vocal
individual’s voice for recognition sounds originate. To produce speech,
purposes. The speaker recognition these components work in combination
process relies on features influenced by with the physical movement of the jaw,
both the physical structure of an tongue, and larynx and resonances in
individual’s vocal tract and the the nasal passages. The acoustic
behavioral characteristics of the patterns of speech come from the
individual. A popular choice for physical characteristics of the airways.
remote authentication due to the Motion of the mouth and
availability of devices for collecting pronunciations are the behavioral
speech samples and its ease of components of this biometric. There
integration, speaker recognition is are two forms of speaker recognition:
different from some other biometric text dependent (constrained mode) and
methods in that speech samples are text independent (unconstrained
captured dynamically or over a period mode). In a system using “text
of time, such as a few seconds. dependent” speech, the individual
Analysis occurs on a model in which presents either a fixed (password) or
changes over time are monitored, prompted (“Please say the numbers
which is similar to other behavioral ‘33-54-63’”) phrase that is
biometrics such as dynamic signature, programmed into the system and can
gait, and keystroke recognition. improve performance especially with
Approach cooperative users. A “text
The physiological component of voice independent” system has no advance
recognition is related to the physical knowledge of the presenter's phrasing
and is much more flexible in situations axis and loudness on the vertical
where the individual submitting the access. The speaker recognition
sample may be unaware of the system analyzes the frequency content
collection or unwilling to cooperate, of the speech and compares
which presents a more difficult characteristics such as the quality,
challenge. Speech samples are duration, intensity dynamics, and pitch
waveforms with time on the horizontal of the signal.

Figure: Voice Sample: The voice input signal (top of image) shows the input loudness with
respect to the time domain. The lower image (blue) depicts the spectral information of the
voice signal. This information is plotted by displaying the time versus the frequency
variations

Vascular Pattern Recognition


Introduction determined that the vascular pattern of
Vascular Pattern Recognition, also the human body is unique to a specific
commonly referred to as Vein Pattern individual and does not change as
Authentication, is a fairly new people age. Claims for the technology
biometric in terms of installed systems. include that it:
Using near-infrared light, reflected or
transmitted images of blood vessels of • Is difficult to forge —
a hand or finger are derived and used Vascular patterns are difficult
for personal recognition. Different to recreate because they are
vendors use different parts of the hand, inside the hand and, for some
palms, or fingers, but rely on a similar approaches; blood needs to
methodology. Researchers have flow to register an image.
• Is contact-less — Users do not hand. Due to the difference in
touch the sensing surface, absorbance of blood vessels and other
which addresses hygiene tissues, the reflected near-infrared rays
concerns and improves user produce an image on the sensor. The
acceptance. image is digitized and further
• Has many and varied uses — processed by image processing

It is deployed in ATMs, techniques producing the extracted

hospitals, and universities in vascular pattern. From the extracted

Japan. Applications include ID vascular pattern, various feature data

verification, high security such as vessel branching points, vessel

physical access control, high thickness, and branching angles are

security network data access, extracted and stored as the template.

and POS access control.


• Is capable of 1:1 and 1: many
matching — Users’ vascular Vascular pattern in fingers

patterns are matched against The basic principle of this technology

personalized ID cards/smart is shown in Figures 1 & 2. Near-

cards or against a database of infrared rays generated from a bank of

many scanned vascular LEDs penetrate the finger or hand and

patterns. are absorbed by the hemoglobin in the


blood. The areas in which the rays are
absorbed (i.e., veins) appear as dark

APPROACH areas similar to a shadow in an image

Vascular pattern in the back of taken by a Charge-Coupled Device

hands (CCD) camera. Image processing can

Near-infrared rays generated from a then construct a vein pattern from the

bank of light emitting diodes (LEDs) captured image. Next this pattern is

penetrate the skin of the back of the digitized and compressed so that it can
be registered as a template.
Figure 1. Transmittance Images of a Hand. Figure 2. Principle of Transmittance Imaging.

Conclusion

Biometrics is a powerful tool to secure would be safe there.


our networks such as banks, offices
At this point, however,
and personal data. It provides high
biometric systems don't have the
security and reduces the incidence of
capability to store and catalog
unauthorized access in sensitive areas.
information about everyone in the
But as no technology is foolproof there
world. Most store a minimal amount of
are some loop holes in this technology
information about a relatively small
which have yet to be covered.
number of users. They don't generally
Some people object to
store a recording or real-life
biometrics for cultural or religious
representation of a person's traits --
reasons. Others imagine a world in
they convert the data into a code. Most
which cameras identify and track them
systems also work in only in the one
as they walk down the street, following
specific place where they're located,
their activities and buying patterns
like an office building or hospital. The
without their consent. They wonder
information in one system isn't
whether companies will sell biometric
necessarily compatible with others,
data the way they sell email addresses
although several organizations are
and phone numbers. People may also
trying to standardize biometric data.
wonder whether a huge database will
exist somewhere that contains vital However above all these advantages

information about everyone in the and disadvantages this technology will

world, and whether that information be creating ripples in the field of


security and privacy.

You might also like