5. Intelligent Human Recognition Systems

collarlimabeansΑσφάλεια

23 Φεβ 2014 (πριν από 3 χρόνια και 1 μήνα)

53 εμφανίσεις

Intelligent Human
Recognition Systems

Dr. C. N. Ravi Kumar

Professor & Head

Dept.of Computer Science & Engg.

S.J.C.E,Mysore

By

What is biometric technologies?


Biometric technologies are essentially a
Pattern Recognition systems which
refers to the "automated methods of
identifying or authenticating the identity
of a living person based on a
physiological or behavioral
characteristic."


Need for Biometrics


PINS (Personal Identification Numbers) and
Passwords may be forgotten.


Token based method of identification like
passports and driver’s licenses may be


Stolen,


Forged


or


Lost.


Some typical biometrics


Primarily Physical Features


Hand based


Fingerprint or fingerscan


Hand geometry


Face/eye


Facial recognition


Retinal scans


Iris scans


Strong Behavioral Component


Voice recognition (verification)


Signature recognition (dynamics)


Key stroke scan

Voice Recognition

Some Biometrics with Reduced
Commercial Viability.


Ear Shape


Odour (human scent)


Vein Scan


Finger geometry


Gait recognition (manner of walking)


Foot Geometry or Foot impression.

Storage Requirements are Higher


“Size” of the template as stored can be quite large
in comparison with a password


Some typical template sizes:


Fingerscan: 250
-

1000 bytes


Hand geometry: 9
-
20 bytes


Iris: 512 bytes


Retina: 96 bytes




Customer Acceptance and Convenience


Does reduce reliance on customer memory



Often considered more convenient than a smartcard



Enrollment style and use of biometrics are easy



Most of the biometrics are fairly stable except for accident
or disease



They are more reliable and accurate



Fingerprint Recognition


Scan takes about 1 second


Size roughly 250
-
1000 bytes for fingerprints (overall)



Technically, most commercial
used Biometric is finger scanning


Finger print technology
captures a representation of the
finger; it involves storing the
image of the finger and
comparing.


http://www.finger
-
scan.com/finger
-
scan_technology.htm


Minutiae matching



Ridge endings



Ridge bifurcation's



Dots



Islands

Fingerprint scanning (
Minutiae based approach
)

Fingerprint Scanners


The technology isn’t
very expensive
anymore
-

scanning
devices can be
purchased for lower
prices.

Combined
Systems

Euclidian Distance Measure:

A Boon for all Biometric
Approaches

2
1
2
2
1
2
12
)
(
)
(
y
y
x
x
d




Speed and Accuracy Claims


Most manufacturers claim high accuracy


False acceptance rates < 1 in 100,000


False rejection rates < 1 in 100


Real
-
time speeds


<1 second to recognize


<5 seconds to enroll


Hand Geometry

Hand Scans


Fairly Complex in usage thus calls for greater discipline
from user's side.


Not the most accurate but not bad; since hands tend to
be similar, it doesn’t do well in a “discovery”
(“identify”) mode


Storage requirements about 9
-
20 bytes


Usually a specialized reader device to measure aspects
such as length, width, thickness, and surface area of the
hand and fingers


Implementation is expensive since it involves high
-
resolution 3
-
D imaging.


How does it work


The camera captures an image of the hand, with the
help of a mirror. The silhouette of the hand is
extracted, and some geometrical characteristics stored.

( See Jain et al.
A Prototype
Hand

Geometry
-
based

Verification

System

)

Applications

BenGurion Airport
-

Hand Geometry

INSPASS
-

Hand Geometry
see
INS Passenger
Accelerated Service System


Facial Scan


Facial recognition is the most natural means of biometric
identification.


Two recognition systems


Holistic features:


PCA (Eigen Faces)


LDA (Linear Discriminant Analysis)




Combination of PCA and LDA.


Local features:


Graph matching


Dynamic link matching


Derivative of the above two.


Neural Network Mapping Technology.

Correlation :

2
2
2
)
,
(
)
(
),
(
)
)(
((
)
,
cov(
E
X
E
x
E
Y
X
E
y
x
x
x
y
x
y
x
y
x
y
x
















)
,
(
y
x

x

Correlation

between 2 random variables X, Y with expected

y

and

and Standard deviations


y
x


values




i
i
i
x
p
X
E
)
(
Where ‘p
i’ is the probabilities of occurrence of
discrete random variables ‘xi’

Examples Iris Images

Iris Scan


Gather data by a camera within 3 feet of eye


512 byte Iris Code represents the visible characteristics of
the eye


It is claimed that the odds of the same Iris Code being
returned by two different people is less than 1 in 10^(52)


Iris Code includes “266 spots” to distinguish between
irises (claim is most other biometrics have between 10
-
60
distinguishing spots)


Iris Code may vary by as much as 25% for a given eye …
but the odds of two different eyes being 75% similar is
said to be 1 in 10^(16), so this seems acceptable


Algorithms used in iris recognition are accurate.

Iris Scan
-

CONTINUED


Iris Scan is based on visible qualities of the iris


A primary visible characteristic is the trabecular
meshwork, a tissue which gives the appearance of
dividing the iris in a radial fashion.


Other visible characteristics include rings, furrows,
freckles and the corona.


Retinal scan

Retinal Scan


Retinal Scan technology is based on the
blood vessel pattern in the retina of the eye
and it provides unique basis for
identification.


Template size small
-

96 bytes


Very accurate representation


Changes likely only from degenerative
diseases


Harder to use than most and requires
participant cooperation


The pattern

of blood vessels that emanate from
the optic nerve and disperse throughout the
retina depends on individuals and never
changes.


No two retinas are the same, even in identical
twins.


Commercial products:
Retinal Technologies




Thermograms



Thermograms requires an infrared camera to
detect the heat patterns of parts of the body that
are unique to every human being (such as the face)


Normally expensive because of the sensors


Useful paper:
Illumination Invariant Face
Recognition Using Thermal Infrared Imagery

(Solikinski & als)

Ear Scan

Ear Scan
-

continued


Surface is smaller (approximately 1/25 or
1/20 of that of space) which allows working
with images of reduced spatial resolution.


More uniform distribution of colour, so that
all information is conserved.


Changes in the shape of ear is slower when
compared to ear.

Biometric methodology

Sample Capture

Feature Extraction

Template Storage

Template comparison & matching

Applications


ATM machine use


Workstations and network access use


Travel and tourism


Internet transactions : E
-
commerce


Public identity cards : Ration cards,Driving Licenses


voter verification


Day care: verify identity of individuals picking up
children


Residential Security


Banking and many more


Role of Smartcard in biometric
system

The combination of a smart card and
biometric logon delivers a two
-
factor level
of security.

Typing



Now a days a biometrics cast its offering into the market
with BioPassword LogOn for Windows NT. The
client/server biometrics application recognizes a user's
typing pattern and uses it to authenticate them to the
network. The software uses a mathematical algorithm to
record pressure, speed, and rhythm as a user types their
user name and password. The typing pattern is compared
against a template created when the software is initially
installed”

Controversies


It's easy to steal a biometric after the measurement is
taken. Once someone steals your biometric, it remains
stolen for life; there's no getting back to a secure situation.



It doesn't handle failure very well.



Not all biometrics are 100% match.


Challenges in Biometrics


Rotation


Tilted


Variation in Intensity


Caemophlage


Occlusion


Rotation

40
0

90
0


Tilted





Variation in Intensity

Caemophlage

Caemophlage

Caemophlage


Occlusion


35%

50%

50%

50%

15%


Complicated Cases in Occluded Images

Conclusion


Biometric systems are not flawless.




One way to enhance the technology is to employ two or
more methodologies.


BIBLIOGRAPHY


Perceptions of Biometrics
-

www.biometricgroup.com.


Association for Biometrics
-

www.afb.org.uk
.


The international Biometric Industry Association
-

www.ibia.org.


BSI (German information security agency)
-

www.bsi.de


The Bio API Consortium
-

www.bioapi.org


The extranet for security professionals
-

www.xsp.org

Questions?

THANK YOU