Towards next generation biometric systems

1
Towards next generation biometric systems

• Marek Rejman-Greene
• Biometric Centre of Expertise
• EBF Research Seminar 2 October 2007

2
Biometric Centre of Expertise

• Advise Home Office (Ministry of Justice)
  programmes and projects
• Work to align programmes
• Support assurance initiatives
• Increase awareness of optimal use of technologies
• Anticipate the future, assessing the threats and
  advising on opportunities
• Outreach engage with academia and the
  commercial sector

3
Multi-modal/Multi-biometric systems

• Benefits
• Accuracy and robustness against degradation
• Efficiency
• Usability adaptability to user capabilities
  preferences
• Security
• Challenges
• Complexity
• Processing / operational time
• Usability - potentially low affordance system
• Standardisation - specification, testing,
  conformance

4
Opportunities

• 2- and 10-finger verification of identity
• US-VISIT, seafarers card
• combined face and stand-off iris units
• advanced fingerprint imagers
• multi-wavelength and multi-polarisation
• contactless/contact-limited 10-print
• interoperable registered traveller schemes
• exception cases - deployment of specialised
  functionality

5
ISO Technical Report 24722

• Multi-Modal and Other Multi-Biometric Fusion
• Taxonomy and definitions
• Multimodal (face, fingerprint)
• Multi-algorithmic
• Multi-instance (more than one finger presented)
• Multi-sensorial (for fingerprints optical TIR,
  silicon, ultrasonic)
• Multi-presentational (rolled and plain
  fingerprints)
• simultaneous and sequential
• Options for fusion at different levels of
  combination
• impact of correlation, quality
• Prospects for standardisation
• likely first for score fusion methods
  (normalisation and fusion methods)

6
Generic single mode system
Feature Extraction
SCORE
MATCH NO MATCH
Sample
Matching
Decision
references
TR 24722
7
references2
decision-level fusion
Feature Extraction2
SCORE
Sample2
Matching2
Decision2
MATCH NO MATCH

Feature Extraction1
SCORE
Sample1
Matching1
Decision1
FUSION (weighted on quality?)
references1
TR 24722
8
references2
decision-level fusion (human comparison) face
(OToole et al, 2007)
SCORE
Sample2
Human
Decision2
Process
MATCH NO MATCH

Feature Extraction1
SCORE
Sample1
Matching1
Decision1
FUSION
references1
9
references2
score-level fusion
Feature Extraction2
Sample2
Matching2
SCORE
MATCH NO MATCH
Decision

Feature Extraction1
Sample1
Matching1
SCORE FUSION (normalised)
references1
TR 24722
10
references2
score-level fusion with (continual) feedback
Feature Extraction2
Sample2
Matching2
SCORE
MATCH POSSIBLE NO MATCH
LOW QUALITY
Decision

Feature Extraction1
Sample1
Matching1
SCORE FUSION (normalised)
HIGH QUALITY
ALTERNATE/ 3rd MODALITY
references1
11
Research

• testing methodologies
• large scale database collection under realistic
  scenarios
• correlation effects
• application to end-to-end systems
• optimisation (cost, time) - security strength
• systems adaptive to ageing, demographics and
  environmental changes
• application to support for privacy and biometric
  encryption
• management of complexity intelligent autonomous
  agents
• specification, interoperability, standardisation,
  conformance
• .

12
System Architecture
University of Kent
13
Case Study
Physician
Patient
Interface
Directory
Access
Agent
Agent
Agent
Biometric
Module
Server registers services
available with directory
A
agent
Fusion
1
Agent
University of Kent
Server
Agent
14
University of Kent
15
Marek Rejman-Greene Senior Biometrics Advisor
16
Feature Extraction
SCORE
MATCH NO MATCH
Sample
Matching
Decision
references
TR 24722