BIOMETRIC DATA MINING

1
Intelligent Biometric Technologies and
Applications
- Simulation and Systems
Patrick Wang,IAPR Fellow , Svetlana Yanushkevich
On leave from Northeastern University, Boston,
MA, USA
2
What are biometric technologies?
Two Al Qaeda suspects were recently taken into
custody by U.S. immigration authorities as they
tried to enter the United States after their
fingerprints were matched with ones lifted by
U.S. military officials from documents found in
caves in Afghanistan.
3
What are Biometric Technologies?
Biometric technologies are automated methods of
recognizing a person based on the acquired
physiological or behavioral characteristics
Facial Scan
Keystroke Scan
Hand Scan
11.4
0.3
10.0
Middleware
12.4
Iris Scan
7.3
Finger Scan
Voice Scan
Signature Scan
52.1
4.1
2.4
Percentage of usage (Source International
biometric group)
4
Biometrics Analysis vs synthesis
Testing biometric systems, simulation for HQP
training and improved security requires synthesis
Recognizing, or authentification of a person is
based on analysis

Analysis
Synthesis

• Inverse methods in analysis
• generation of variations (morphed/distorted
  signals) at the learning stage for classifying
  and pattern recognition software

• Inverse Biometrics
• generation of synthetic biometric data or
  encoding (distorting) data

5
AnalysisPattern Recognition Concept
Sensors
Extractors Image- and signal- pro. algo.
Classifiers
Negotiator
Threshold
Decision Match, Non-match, Inconclusive
Biometrics Voice, signature acoustics, face,
fingerprint, iris, hand geometry, etc
Data Rep. Audio signal, image, infrared image
Feature Vectors
Scores
Enrolment
Training
Submission
6
How it works Fingerprint Extraction and Matching
7
How it works IrisCode
8
How it works Facial recognition
Principal Component Analysis
9
Authentication Applications

• Cell phones, Laptops, Work Stations, PDA
  Handheld device set.
• 2. Door, Car, Garage Access
• 3. ATM Access, Smart card

Image Source http//www.voice-security.com/Apps.
html
10
Our lab project Screening and early warning
system
Screening discipline of observation
Person
Visible band camera
Visible band camera
IR band camera
Phase 2 (specified person)
Officer
IR band camera
Phase 1 (unknown person)
Voice analyzer
Dialog supporting
Forming the persons file
Processing preliminary data
On-line processing
Phase 3 Decision making
Global database
Expert system
11
Laboratory experiments
12
Thermal Imagingemotion recognition
Thermal image of subject at the beginning of
answering the question Do you have that stolen
20 on you right now?
Visualization of the blood flow rate (c) from the
upper rectangle of (a)
Visualization of the blood flow rate (d) from
(b). The difference is significant
Thermal image of subject at the end of answering
the question (ref. I. Pavlidis work)
13
Synthesis in biometrics

• Biometric Synthesis rendering biometric
  phenomena from their corresponding computer
  models
• Drawing parallels Computer Vision (analysis) and
  Computer Graphics (synthesis)

14
Where do we need synthetic biometrics?

• Traditional application speech synthesis
• Another application signature forgery
• Analysis-through-synthesis for improving the
  verification algorithms
• Testing biometric devices and systems
• Designing simulators for HQP training systems
• Sensing in robotics

15
Example Face Synthesis
Mesh-based model SDK Mesh/texture modeling tool
FaceGeneration (Singular Inversions Inc.)
Customizing morphs modification of vertex
coordinates to perform the morphs (smile, raise
eyebrow, surprise etc) Current project of the
Biometric Technologies Lab
16
Example Face synthesis
Face capturing
Fitting points
00010010010100110100100100100101100100100010000100
10110100100101001001001000
File (mesh/colour)
3D Face model
17
Example modeling the appearance
original
disguised
aged
drunk
badly lit
18
Example Modeling Facial Nerve Disorders

• Picture courtesy of Dr. Toth (Neurosciences at
  the UofC Medical Clinic)

• Facial Nerve Disorders
• Left Bells Palsy (paralysis of one side of face
  )
• Ptosis drooping of eyelids
• Mobius Syndrome limited facial muscle movements

The project has been implemented in the
Biometric Lab (undergraduate project)
19
Example other models
3D from 2D Intensity -gt Filtering -gt Magnitude
of the surface gradient -gt Profile generation -gt
Surface gradient -gt Filtering current project
of the Biometric Lab
20
Example synthetic iris
Real iris image
Synthetic iris pattern
Iris Pattern Generator project has been
implemented using Matlab in the the Biometric Lab
(A.Widmer, O.Mulhauser)
21
Example Synthetic iris
Hand-painted patterns
Distance transform
Wavelet transform
Fourier transform polar transform
Voronoi diagram
22
Application of synthetic irises
Testing the best iris recognition system has
been tested using only 4258 different iris images
(out of 12 billion irises of the world
population). Solution synthesis of iris patterns
Panasonic Authenticam
- Medicine patterns for the eye prostheses are
currently handmade
23
Example fingerprint synthesis
Real Fingerprint
Synthetic fingerprint

• Known implementations SFinGe software
  (University of Cesena, Italy)
• Another project for generating fingerprints has
  been implemented in the Biometric Technologies
  lab (P. Wang)

24
Application of synthetic fingerprints
- Creation of databases of prints for RD as well
as training. Different pressures, roles,
moisture, oiliness, cuts, folds. No privacy
concerns.
25
Application of synthetic fingerprints
Testing of fingerprint readers
Normal Test Finger Index Middle Ring
Small Thumb FAR 0 0.00002 0.00002
0.00001 0.00005 FRR 0 0.038 0.200
0.353 0.038
Dry Test Finger Index Middle Ring
Small Thumb FAR 0.000007 0 0.000014
0.000022 0.000029 FRR 0.025 0.150 0.325
0.425 0.025
Wet Test Finger Index Middle Ring
Small Thumb FAR 0.000014 0.000061
0.0000072 0.000014 0.000057 FRR 0.575 0.500
0.550 0.625 0.350
Heinmann optical fingerprint reader, 3550
fingerprints (200 real 3350 synthetic) 10
fingerprints per volunteer synthetic prints a
test base of 355 volunteers this provides 95
confidence of an error rate of 3/3550.845
26
Example Forged signature
3D on-line capture with a Wacom Tablet
Synthetic Signature (joint project with Baker
University, courtesy Dr. D. Popel)
27
An Overall Interactive Intelligent
Learning/Recognition System
28
Basic Concept of Measurement
29
Example of Ambiguity
30
References
Yanushkevich S, Stoika A., Shmerko V., Popel D.,
Biometric Inverse Problems, CRC
Press/TaylorFrancis, 2005 Yanushkevich S., Wang
P., Srihari S., Gavrilova M, Image Pattern
Recognition Synthesis versus Analysis in
Biometrics, World Scientific Publishers, 2007 (to
appear) P. Wang, Recent Developments and
Applications of Pattern Recognition and
Biometrics, in Image Pattern Recognition
Synthesis versus Analysis in Biometrics, World
Scientific Publishers, 2007 (to appear)