Application of Multibiometric Fusion Techniques Rick Lazarick, Chief Scientist CSC GSS IDENTITY LABS - PowerPoint PPT Presentation

Loading...

PPT – Application of Multibiometric Fusion Techniques Rick Lazarick, Chief Scientist CSC GSS IDENTITY LABS PowerPoint presentation | free to view - id: 205824-MmFhY



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Application of Multibiometric Fusion Techniques Rick Lazarick, Chief Scientist CSC GSS IDENTITY LABS

Description:

the automated recognition of individuals based on their biological or behavioral ... (Not meant to describe any existing system) ... – PowerPoint PPT presentation

Number of Views:552
Avg rating:3.0/5.0

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Application of Multibiometric Fusion Techniques Rick Lazarick, Chief Scientist CSC GSS IDENTITY LABS


1
Application of Multibiometric Fusion Techniques
Rick Lazarick, Chief Scientist
CSC - GSS IDENTITY LABS

Panel Multi-Modal Biometrics
6th International Public Safety/Counterterrorism
Conference April 23, 2007
2
Presentation Outline
Why Multibiometrics? Types of multibiometric
fusion Normalization and fusion techniques Pros
and cons of multibiometrics Determining the
appropriated degree of complexity
3
Challenges
  • Challenges for Biometric Applications
  • Enrollment rate vs. effort ()
  • Selection of biometric modality (s)
  • Policy decisions
  • Resolving false rejections
  • Secondary procedures and policy
  • Throughput (at high volume entry points)
  • Harsh environment
  • Balancing security and convenience
  • Imposter frequency and sophistication

4
Multibiometrics to the Rescue?
The intelligent application of multibiometrics
can help! So, what is multibiometrics?
5
WHY Multibiometrics?
  • To Reduce (some or all of)
  • False acceptance rate
  • False rejection rate
  • Failure to enroll rate
  • Failure to acquire rate
  • Susceptibility to spoofing

6
WHY NOT Multibiometrics?
  • Possible Drawbacks
  • Sensor acquisition cost
  • Enrollment time/cost
  • Transit times
  • Need for a priori data
  • System development or complexity

7
Fusion 101
Multibiometrics the automated recognition of
individuals based on their biological or
behavioral characteristics and involving the use
of biometric fusion Approaches to Multibiometric
Fusion Multimodal Multialgorithmic Multiinstanc
e Multisensorial Hybrid Based on ISO TR 24722
8
Fusion 101
Levels of Fusion Decision Score Feature Sample
9
Fusion 101
Score level fusion Recently most studied
technique Often requires normalization Many
fusion methods defined
10
Normalization Techniques
  • Normalization required to bring dissimilarly
    scaled matching scores into a common basis
  • Many techniques some require significant a
    priori data
  • Techniques listed in ISO Technical Report
    Multimodal and other Multibiometric Fusion (TR
    24722)

11
Fusion Methods
  • Score level fusion techniques are numerous and
    varied in complexity and performance
  • Some techniques have underlying assumptions
    and/or require significant a priori data
  • Techniques listed in ISO Technical Report
    Multimodal and other Multibiometric Fusion (TR
    24722)

12
Hypothetical Examples
Multibiometric Applications
  • (Not meant to describe any existing system)
  • Used here to illustrate the language and to
    explore strengths and weaknesses

13
Example 1- Multimodal, decision level fusion with
sequential sampling
Hypothetical Examples
Application Attended Physical Access
Control Description Fingerprint and Iris, OR
logic, fingerprint first
  • Disadvantages
  • Additional hardware costs (two sensor systems)
  • Additional enrollment time
  • Increased potential for false accepts
  • Advantages
  • High user enrollment rate (either modality
    policy)
  • High throughput potential (finger first)
  • Low false reject rate

14
Example 2- Multiinstance, score level fusion with
sequential sampling
Hypothetical Examples
Application Unattended Physical (and/or
Logical) Access Control Description Multiple
Fingerprints, single digit sensor, verification
(1-to-1) with Sum Rule Combination Fusion
  • Advantages
  • Low Cost (primary driver)
  • Low false reject rate with high security
  • Option for higher spoof resistance (using
    query/response different fingers)
  • Disadvantages
  • Longer usage time (multiple sequential samples)
  • Some users unable to enroll

15
Example 3- Multisensorial, multialgorithmic,
hybrid fusion with simultaneous sampling
Hypothetical Examples
Application Token-less Identification for
Privileged Access Description Dual sensor Face
Recognition (2-D and 3-D), each with multiple
matchers, individualized weighted sum rule and
voting scheme
  • Disadvantages
  • Lengthy enrollment time
  • Very complex logic requiring tuning
  • Costly hardware
  • Advantages
  • User satisfaction (no token to forget, no
    contact)
  • Low failure to enroll rate
  • Potentially very high accuracy identification
    rates

Note potential future application of face and
long range iris
16
Analysis of Benefits
  • Error Rate Reduction
  • The following is an illustration of analysis
    results using public domain matchers and data
  • Compare single and multiple modalities
  • Face and two Fingerprints
  • Unimodal false reject range 14 26
  • Optimum fusion false reject 2 3
  • Conclusion dramatic reduction in error rate
    achievable
  • Values quoted at false accept 1 in 10,000

17
MINEX
  • Multi-instance fingerprint verification
    (Example)
  • MINEX results for multi-instance (L and R index)
  • Score-level fusion using simple sum
  • (no normalization required)
  • MINEX Minutiae Interoperability Exchange Test

18
MINEX
  • Multi-instance fingerprint verification
    (Findings)
  • Improvement in FMNR for all algorithms
  • Best algorithms have gt10-fold reduction in error
  • Average algorithms have 5-fold reduction in
    error
  • Possible inconvenience factor- two
    presentations?

19
Emerging Multibiometric Concepts
Face Iris Single image high resolution
camera Perception of public acceptability Iris
Retina Single presentation (simultaneous
capture) Significantly cooperative user Palm
print Hand Geometry Single presentation
(simultaneous capture) 2-D optical scanner (low
cost)
20
Multibiometric Cost-Benefit Trade
  • Key decision considerations
  • Is the benefit of implementing a multibiometric
    system proportional to the investment?
  • How much is enough?
  • Research consideration
  • How to measure the Return on Investment (ROI)?

21
(No Transcript)
22
Experience. Results.
Rick Lazarick, Chief Scientist Global Security
Solutions Identity Labs Computer Sciences
Corporation
About PowerShow.com