Part 1 Ideas for themes/topics/misconceptions
1. Biometrics – Reality and Myths
2. Biometrics in Real Life
3. Biometrics –privacy
Hollywood Face Recognition
Common misconceptions
– 100% match to any image at
any angle
– Instantly recognize any
person
– Tied into a “super database”
that knows who everyone is
– Available to and in use by
law enforcement
Movie scene
(downloaded from www
pubic domain)
Hollywood DNA
Misconceptions
– Access to a super
database that has
everyone’s DNA
– Automatically and
rapidly processes a
sample
Movie scene
(Pubic domain)
Hollywood Fingerprints
Common misconceptions
– instant match
– Only one fingerprint is
enough
– Available to use at any
location
Screenshot from “Man in black” movie
(Pubic domain)
Face Recognition Today
• Today’s Reality
– Affected by lighting, angle,
quality of captured image
– Requires a “high-end”
computer for real-time face
capture/processing
– Many are stand-alone
systems
– Being evaluated, not
deployed
Ft. Lauderdale Airport, Florida
IdentiFace system interface
Face Recognition Today
• Today’s Reality
– Varying confidence of match
depending on application
– Multiple unique and
proprietary image formats
make sharing hard
– Intelligence images not
available to local law
enforcement or corrections
– Data sharing across jurisdictions
is a problem
100 known images in the database
Identiface
Face Recognition Today
• Face Recognition Vendor Test 2002
and 2006 provides independent
government evaluations of
commercially available and mature
prototype face recognition systems.
• Results available at
http://www.itl.nist.gov/iad/894.03/f
ace/face.html
•
•
•
•
FRVT 2002 and 2006 evaluated
performance on:
High resolution still imagery (5 to
6 mega-pixels)
3D facial scans
Multi-sample still facial imagery
Pre-processing algorithms that
compensate for pose and
illumination
FpVTE Fingerprint Vendor Technology
Evaluation
• The Fingerprint Vendor Technology Evaluation (FpVTE) 2003 is an
independently administered technology evaluation of fingerprint
matching, identification, and verification systems.
• Assessed the capability of 18 vendors fingerprint systems to meet
requirements for large-scale and small-scale real applications.
• Consists of multiple tests performed with combinations of fingers and
different types and qualities of operational fingerprints
• Conducted by the National Institute of Standards & Technology (NIST)
between October and November 2003 on behalf of U.S. Department of
Justice.
• Report made public in June 2004 at http://FpVTE.nist.gov
FpVTE Fingerprint Vendor Technology
Evaluation
Some of the Results
• Systems that performed most accurately were developed by
NEC, SAGEM, and Cogent
• The variables that had the largest effect on system accuracy
were the number of fingers used and fingerprint quality.
• Different systems were distinguished by how they performed
across the spectrum from good to bad (performance
separation was really on “bad” quality).
Applications: Biometrics in Schools
•
National News Reports
•
•
Interior System
Exterior System
Eleven, Single-Eye LG Electronics
IrisAccess 2200 Iris Recognition
Cameras were Evaluated
– 6 cameras within closed areas in
3 schools
– 5 cameras were located
outdoors with fabricated
protective closures
Unsuccessful attempts mostly due to
camera capture errors (16%) and
access attempts by unknown users
(5.8%)
Issues remaining:
– Tailgating (accepted users
holding door open for others)
– Ability to Capture Iris Outdoors
(lighting conditions)
Applications: Biometrics in Correction
Facilities
•
Demonstration and Assessment of
Facial Recognition Technology at
Prince George’s County Correctional
Facility (Visionics corporation)
• Visionics (now Identix) system
installed based on results of FRVT
2000
• Required re-work of room
lighting, addition of camera lights,
and training of staff and system
users.
• Interfaced with Staff and
Volunteer Access Control System
to verify identity of staff and
volunteers upon entry and exit
from the facility
• Augments manned access control
station
PRIVACY: Video surveillance
When you walk into a
building from a parking lot?
When you shop at your
favorite store?
Go to your bank?
Privacy: Video surveillance
When you buy gas for your
car?
Pay at a toll booth?
Video surveillance is a
daily fact of life.
Current motivation is
mostly to avoid theft in
commerce.
Summary:
• Biometric Technology is concerned with representation,
storage, matching, synthesis and visualization of biometric
information.
• Tremendous advance has been achieved over the last few
years in both fundamental theoretical development,
matching and synthesis, as well as biometric hardware and
software products.
• Many misconceptions remain, and are supported by mass
media and newspapers
• There are real privacy concerns with video surveillance
technology, which is not biometrics.
Part 2 Ideas for biometric theme – definitions,
variety of resources
1. Biometrics Identifiers - classification
2. Biometric market share
3. Biometric – how to make choice
Biometric identifiers
From G. Bromba research
Biometric Market Share
Comparison of biometric techniques
Palm
Hand vein
Facial Thermogram
Ear print
Retina
Which Biometric is the Best?
•
•
•
•
•
Universality (everyone should have this trait)
Uniqueness (everyone has a different value)
Permanence (should be invariant with time)
Collectability (can be measured quantitatively)
Performance (achievable recognition accuracy, resources
required, operating environment)
• Acceptability (are people willing to accept it?)
• Circumvention (how easily can it be spoofed?)
Selecting a Biometric
Selecting the ‘right’ biometric is a complicated problem
that involves more factors than just accuracy. It depends
on cost, error rates, computational speed, reliability,
privacy and easy of use.
Part 3 Historical artifacts
1. History of fingerprints
2. First automated computer systems
3. Historical artifacts
4. Fingerprint devices in real life
5. Optional – sensing devices
6. Face recognition – methods, illusions, image
synthesis
7. Other biometric devices
Cartoon (copyrighted)
History of fingerprints
• Use of fingerprints to associate a person with
an event or transaction can be traced to
ancient China, Babylonia and Assyria as early
as 6,000 BC.
28
Archaelogical remains
29
History of fingerprints
• 1750 B.C.- people in Babylon used
fingerprints to sign their identity on clay
tablets
• 300 B.C.-Emperors of China used
personalized clay seals
• In 1686, Marcello Malpighi, an anatomy
professor at the University of Bologna,
wrote in a paper that fingerprints
contained ridges, spirals and loops.
• In 1856, Sir William Herschel, a British
Magistrate in Jungipoor, India, used
fingerprints (actually palmprints) to certify
native contracts.
30
History of fingerprints
• During the 1870s, Dr. Henry Faulds, a British
surgeon in Japan, after noticing finger marks
on ancient pottery, studied fingerprints,
recognized the potential for identification,
and devised a method for classifying
fingerprint patterns.
• 1880 -Faulds published an article in
"Nature," discussing fingerprints as a means
of personal identification. He is also credited
with the first fingerprint identification of a
greasy fingerprint left on an alcohol bottle.
31
32
History of fingerprints
• In 1880, Alphonse Bertillon, a Paris police department
employee and son of an anthropologist, developed a system
of anthropometry as a means for classifying criminals and
used this system to identify recidivists.
Anthropometry (a system of cataloging an individual's body
measurements such as height, weight, lengths of arm, leg,
index finger etc.) was shown to fail in a famous case at
Leavenworth Prison, where two prisoners, both named
William West, were found to have nearly identical
measurements even though they claimed not to be
biologically related.
33
Alphonse Bertillon
Alphonse Bertillon
34
History of fingerprints
• Francis Galton, an anthropologist, began a systematic study of
fingerprints as a means of identification in the 1880s.
In 1892, he published the first book on fingerprints.
• In 1897, Sir Edward Henry, a British police officer in India,
established a modified fingerprint classification system using
Galton's observations. This system was ultimately adopted by
Scotland Yard in 1901 and is still used in many Englishspeaking countries.
35
Historical Overview
• Henry Faulds, William Herschel and Sir Francis Galton
proposed quantitative identification through fingerprint
and facial measurements in the 1880s.
• Edmond Locard introduced using
biometrics in forensic identification
in 1920s.
36
History of fingerprints
• 1924-an act of U.S. Congress established the Identification
Division of the FBI (Federal Bureau of Investigation) with a
database of 810 000 fingerprint cards
• Most of the early fingerprint identification systems were put
into place in major metropolitan areas or as national
repositories. Juan Vucetich established a fingerprint file
system in Argentina in 1891, followed by Sir Edward Henry in
1901 at Scotland Yard in England.
37
Classification: Manual Card Files
In the Henry classification system, numerical weights are
assigned to fingers with a whorl pattern. A bin number, based
on the sum of the weights for the right hand and sum of the
weights for the left hand is computed to generate 1,024
possible bins. Letter symbols are assigned to fingers: capital
letters to the index fingers and lower-case letters to other
fingers.
These are combined with the numeric code to further
subdivide the 1,024 bins. Each of these pattern groupings
defines bins into which fingerprint cards with the same
pattern group are placed.
A bin might be a folder in a file drawer or several file drawers if it
contains a common pattern group and the file is large.
38
Classification
• Fingerprint patterns comprise
of loops (left or right), whorls
and arches. The patterns are
differentiated based on the
presence of zero, one or two
delta regions.
A delta region is defined by a
tri-radial ridge direction at a
point. Arch patterns have no
delta, loops have one delta,
and whorls have two deltas.
Some of the common fingerprint
types. The core points are marked
with solid white circles while the
delta points are marked with solid
black circles.
39
Matching
Fingerprint matching prior to automation involved the manual
examination of the so-called Galton details (ridge endings,
bifurcations, lakes, islands, pores etc. collectively known as
"minutiae").
Prior to the late 1960s, neither the available computer
systems that could display fingerprint images for comparison
were affordable, nor a significant number of digital fingerprint
images were available for display. Consequently, the
comparison was manual, requiring a magnification glass for
comparing the features of the many candidate prints
manually retrieved from the database files.
40
Early Automation Efforts
• US NBS/NIST Research: In the mid-1960s, the National
Institute of Standards and Technology (NIST) initiated
several research projects to automate the fingerprint
identification process. The efforts were supported by the
Federal Bureau of Investigation (FBI) as part of an initiative
to automate many of the processes in the Bureau.
• Royal Canadian Police: By the mid-1960s, the fingerprint
collection of the Royal Canadian Mounted Police (RCMP)
had grown to over a million tenprint records. The videofile system was operational until the mid-1970s, when
the RCMP installed the first automated fingerprint
identification system (AFIS).
41
Early Automation Efforts
• FBI: In the USA, at about the same time that the RCMP and
the UK Home Office were looking for automation
technologies, the FBI was investigating the possibilities for
automating the fingerprint identification operations. In the
mid-1960s, the FBI signed research contracts with 3
companies to build a working prototype for scanning FBI
fingerprint cards, completed by the early 1970s.
• United Kingdom: In the UK, over about the same time-scale as
the FBI, the Home Office was working within its own Scientific
Research and Development Branch (SRDB).
• Japan: The Japanese National Police (JNP) who had a
fingerprint file of over six million records, also initiated study
of the automation possibilities.
42
Further Development
•
During the 1970s, the FBI contracted with a number of organizations as well as
developed their own research organization to manage the numerous projects that
lead the way to the Integrated Automated Fingerprint Identification System (IAFIS).
•
The transition to a large-scale imaging application environment provided
enormous challenges for everyone at that time, but it was especially challenging
for the FBI to implement a system to manage up to 35,000 image-based
transactions per day.
•
The efforts put into AFIS interoperability by NIST under the FBI sponsorship
resulted in an ANSI/NIST standard for data interchange. This standard was initially
crafted in mid-1980, is updated every 5 years and defines data formats for images,
features and text.
43
Early Devices
• The FBI initiated a research program to build an engineering model
of a scanner that could sample an object area of 1.5 X 1.5 in at 500
pixels per inch (DPI), with an effective sampling spot size of 0.0015
in, signal-to-noise ratio (S/N) in excess of 100:1 and digitized to at
least 6 bits (64 gray levels).
• In the late 1960s, these requirements could only be met by a
system that used a cathode ray tube and a precision deflection
system, an array of tubes that measure and reflected light, and
amplifier-digitizer to convert the electrical signal into a digital
value for each pixel.
• There were relatively few scanning devices by the late 1970s that
met the technical characteristics requirements of 500 dpi, a 0.0015
inch effective sample size, greater than 100 S/N noise ratio and a 6
bit dynamic range.
• It ten more years the scan quality standards were set by IAFIS,
which is the current benchmark for scanning of fingerprint images,
44
requiring 200 or more gray levels.
Overview
EarlyHistorical
Computerized
Systems
•
Speaker and fingerprint recognition systems were first AUTOMATED
systems in 1960. The potential for application of this technology to
high-security access control, personal locks and financial transactions
were recognized in the early 1960s.
•
The 1970s saw development and deployment of hand geometry
systems, the start of large-scale testing and increasing interest in
government use of these "automated personal identification"
technologies. There are currently 180 readers used by about 18,000
enrolled users.
•
Retinal and signature verification systems came in the 1980s, followed
by the face systems.
•
Iris recognition systems were developed in the 1990s.
45
Biometrics
and Privacy
Modern examples
Biometric measures can be used in place of a name, Social Security
number or other form of identification to secure anonymous transactions.
Walt Disney World sells season passes to buyers anonymously, then uses
finger geometry to verify that the passes are not being transferred. (Image of pass to
Disney World)
The real fear is that biometric measures will link people to personal data,
or allow movements to be tracked. After all, credit card and phone records
can be used in court to establish a person's activities and movements.
Phone books are public databases linking people to their phone number.
These databases are accessible on the Internet. “Reverse" phone books also exist (a
name from a phone number).
Unlike phone books, databases of biometric measures cannot
generally be reversed to reveal names from measures because biometric
measures, although distinctive, are not unique.
46
Modern Examples
47
Biometrics
and Privacy
Everyday applications
• Samples of these documents can be requested:
• Five US states have electronic fingerprint records of social service
recipients (Arizona, California, Connecticut, New York and Texas).
• Six states (California, Colorado, Georgia, Hawaii, Oklahoma and Texas)
maintain electronic fingerprints of all licensed drivers.
• Nearly all states maintain copies of driver's license and social service
recipient photos.
• FBI and state governments maintain fingerprint databases on convicted
felons and sex offenders.
• Federal government maintains hand geometry records on those who have
voluntarily requested border crossing cards.
48
•
Interesting facts
•
Like telephone and credit card information, biometric databases can be
searched outside of their intended purpose by court order.
•
Unlike credit card, telephone or Social Security numbers, biometric
characteristics change from one measurement to the next.
•
Unlike more common forms of identification, biometric measures contain
no personal information and are more difficult to forge or steal.
•
Biometric measures can be used in place of a name or Social Security
number to secure anonymous transactions.
•
Some biometric measures (face images, voice signals and "latent"
fingerprints left on surfaces) can be taken without a person's knowledge,
but cannot be linked to an identity without a pre-existing database.
•
Searching for personal data based on biometric measures is not as reliable
or efficient as using better identifiers, like legal name or Social Security
number.
•
Biometric measures are not always secret, but are sometimes publicly
observable and cannot be revoked if compromised.
Biometric and Privacy
49
OPTIONAL: Fingerprint Sensing: optical
FTIR
Frustered Total Internal Reflection (FTIR):
The oldest and most used livescan
technique. Reflection allows the
ridges (appear dark) to be
discriminated from the valleys
(appear bright). Because devices
sense a 3D surface, they cannot be
easily deceived by a photograph or
printed image of a fingerprint.
50
Fingerprint Sensing: FTIR with a sheet
prism
51
Optical Fibers
52
Optical Sensors
53
Ultrasound sensors
54
Solid State Sensing
Thermal
Electric Field
Piezoelectric
55
Really rare examples of fingerprints
56
Face Recognition (copyrighted
cartoon)
Imaging
Face tracking in digital cameras
FotoNation Face Tracker
http://www.fotonation.com/index.php?module=product&item=23
Tracking through background
Cha Zhang (Microsoft Research) – my
colleague- uses background
segmentation for face identification and
tracking
Psychological Aspects of Face
Perception and Recognition
Visually Derived Semantic Categories of Faces
Humans can recognize faces along a number of dimensions referred as
• “visually derived semantic categories”: race, sex, and age.
• personality characteristics (visually specified, albeit abstract, categories):
face that looks “generous”, or “extroverted”
An intriguing aspect of this phenomenon is that making such judgments
actually increases human accuracy when compared to making physical
feature-based judgments (e.g., nose size).
In contrast to the information needed to specify facial identity, visually
derived semantic categorizations are based on the features a face shares
with an entire category of faces.
There has been less research on the perception of visually derived semantic
categories than on the face recognition.
Thatcher illusion. The inverted face appears normal. Upright,
however, the configural distortion is evident. The illusion illustrates
the limits of configural processing for atypical views of the face.
Modeling expression
This system (FACS) is
purely descriptive and
includes no inferential
labels.
By converting codes to
EMFACS or similar
systems, face images
may be coded for
emotion-specified
expressions as well as
for more molar categories
of positive or negative
emotion.
64
Computationally-derived information
in images and three dimensional
head models that specifies the
gender of a face.
SUMMARY part 3
Although many recent advances and successes in automatic facial
expression analysis have been achieved, many questions remain
open. Some major ones are:
How do humans correctly recognize facial expressions?
Is it always better to analyze finer levels of expression?
Is there any better way to code facial expressions for computer
systems?
How do we obtain reliable ground truth?
How do we recognize facial expressions in real life?
Ho do we best use the temporal information?
How may we integrate facial expression analysis with other
modalities?
Which role psychology plays in the process?
What impact quality of devices has on the result of
recognition?
Other technologies – signature
authentication
Digitizing Tablet
Iris synthesis
An ocularist's approach to human iris synthesis. Lefohn et. al. 2003.
Used with permission.
Summary:
In BT lab, we are trying to find answer
on those and other questions related to
multi-modal biometric and intelligent
data representation and processing.
Asimo, Honda robot – one example of attempt
to make robot resemble humans in both
appearance and action.