3. Why multimodal biometrics?
• Unimodal biometric systems perform person recognition
based on a single source of biometric information.
• Such systems are often affected by the following
problems:
(1) Noisy sensor data: Noise can be present in the acquired
biometric data mainly due to defective or improperly
maintained sensors.
.
4. Continue
(2) Non-universality: If every individual in the target
population is able to present the biometric trait for recognition,
then the trait is said to be universal. However, not all biometric
traits are truly universal.
people with hand-related disabilities, manual workers with
many cuts and bruises on their fingertips, and people with very
oily or dry fingers
NIST reported 2% people cannot enroll using finger print.
5. Continue
(3) Lack of individuality : Features extracted from
biometric characteristics of different individuals can be quite
similar.
A small proportion of the population can have nearly identical
facial appearance due to genetic factors (e.g., father and son,
identical twins, etc.)
(4) Lack of invariant representation :The biometric
data acquired from a user during verification will not be identical
to the data used for generating the user’s template during
enrollment.
6. Continue
(5) Susceptibility to circumvention:Although it is
very difficult to steal someone’s biometric traits, it is still
possible for an impostor to circumvent a biometric system using
spoofed traits.
Behavioral traits like signature and voice are more susceptible
to such attacks than physiological traits.
TEST TEST FALSE FALSE
PARAMETER REJECT RATE ACCEPT RATE
Fingerprint FVC[2004] 20 years 2% 2%
(average age)
Face FRVT[2002] Varied lighting 10% 1%
Outdoor/indoor
Voice NIST[2000] Text 10-20% 2-5%
Independent
7. Multimodal biometric systems
• Use of multiple biometric indicators for identifying
individuals, known as multimodal biometrics. Combining the
evidence obtained from different modalities using an effective
fusion scheme can significantly improve the overall accuracy
of the biometric system.
• . A multimodal biometric system can reduce the FTE/FTC rates
and provide more resistance against spoofing because it is
difficult to simultaneously spoof multiple biometric sources.
• Four levels of information fusion are possible in a multimodal
biometric system. They are fusion at the sensor level, feature
extraction level, matching score level and decision level.
8. .
Which Biometric Modalities to Fuse?
Voice, Face
Voice, Lip Movement
Voice, Face, Lip Movement
Fingerprint, Face
Fingerprint, Face, Voice
Fingerprint, Face, Hand geometry
Fingerprint, Voice, Hand geometry
Fingerprint, Hand geometry
Facial thermogram, Face
Iris, Face
Palmprint, Hand geometry
Ear, Voice
9. Classification
• Multimodal biometric systems that have been
proposed can be classified based on four parameters,
namely
(1) architecture
(2) sources that provide multiple evidence
(3) level of fusion
(4) methodology used for integrating multiple cues
10. Architecture
• Architecture of a multimodal biometric system refers to the
sequence in which the multiple cues are acquired and
processed.
• Two types
1) serial
2)parallel
• Serial Architecture: In the serial or cascade architecture,
the processing of the modalities takes place sequentially and
the outcome of one modality affects the processing of the
subsequent modalities. Ex: bank ATMs
• Parallel Architecture :In the parallel design, different
modalities operate independently and their results are
combined using an appropriate fusion scheme. Ex: in military
13. Levels of fusion
• Broadly categorized into 2 types
a) fusion prior to matching
b) fusion after matching
• In Fusion prior to matching integration of information can take
place either at the sensor level or at the feature level.
• Sensor level:: Sensor level fusion can be done only if the
multiple cues are either instances of the same biometric trait
obtained from multiple compatible sensors or multiple
instances of the same biometric trait obtained using a single
sensor,ex: 3D model of face.
• In sensor level fusion, the multiple cues must be compatible
and the correspondences between points in the data must be
known in advance.
• It may not be possible to integrate face images obtained from
cameras with different resolutions
14. .
• Feature level: When the feature vectors are homogeneous
(e.g., multiple fingerprint impressions of a user’s finger), a
single resultant feature vector can be calculated as a weighted
average of the individual feature vectors.
• When the feature vectors are non-homogeneous (e.g., feature
vectors of different biometric modalities like face and hand
geometry), we can concatenate them to form a single feature
vector. features vectors must be compatible.
• Integration at the feature level is difficult to achieve in practice
because of the following reasons:
(i) The relationship between the feature spaces of different
biometric systems may not be known.
(ii) Concatenating two feature vectors may result in a feature
vector with very large dimensionality leading to the ‘curse of
dimensionality’ problem.
15. • Fusion after matching: categories into
1) Dynamic classifier selection scheme: chooses
the results of that classifier which is most likely to give the
correct decision for the specific input pattern.
2)Abstract or decision level: can take place when each
biometric matcher individually decides on the best match
based on the input presented to it. Methods like majority
voting,And rule Or rule can be used to arrive at the final
decision.
3)Rank level:When the output of each biometric matcher is a
subset of possible matches sorted in decreasing order of
confidence, the fusion can be done at the rank level.so rank is
assigned from highest to lowest level.
4)Matching score level:
16.
17. Fusion at the Matching Score Level
• Two possible approaches in the context of verification:
– Classification approach: A feature vector is constructed
using the matching scores. Feature vector is classified as
belonging to either genuine or impostor class (e.g., k-Nearest
Neighbor, Decision tree)
– Combination approach: A single scalar score is
generated from multiple matching scores. A classifier
is designed to operate on the new score (e.g.simple sum,
Min score, max score, matcher weighting,user weighting)
Experiments indicate that the combination approach
performs better than the classification approach
19. Score Normalisation
• Scores output by individual matchers:
– Non-homogeneous: distance or similarity
– Ranges may be different; e.g., [0,100] or [0,1000]
– Distributions may be different
• To facilitate fusion:
– Modify the location and scale parameters of score
distributions of individual matchers.
– Apply transformation to scores present in the genuine
impostor overlap region.
20. Normalization Techniques
Min-Max(MM):This method maps the raw scores to the [0, 1]
range . The quantities max(S) and min(S) specify the end points of
the score range:
S:set of all scores for that matcher
s:a raw matching score
Z Score(ZS):
Tanh(TH):
It maps the raw scores to the (0, 1) range
21. • Adaptive(AD): The errors of individual biometric matchers stem
from the overlap of the genuine and impostor score distributions. This
overlap region represented by its center c and its width w. To decrease
the effect of this overlap on the fusion algorithm, an adaptive
normalization procedure apply and aims to increase the separation of the
genuine and impostor distributions, while still mapping the scores to [0,1]
range.
22. Biometric Fusion
• These are of following types
m
n i
represents the normalized score for matcher m (m=1,2…
M ,where M is the number of matchers) applied to user
i (i=1,2…I ,where I is the number of individuals in the database).
Simple Sum(SS):
Min Score(MIS):
Max Score(MAS):
23. continue
• Matcher Weighting(MW): Weights are assigned to the
individual matchers based on their Equal Error Rates (EER’s).
m
Denote the EER of matcher m as e
Where m=1,2….M
Weight associated with matcher m is calculated as
Fused score for user i is calculated as
24.
25. Conclusion
• Though time taken in Multimodal biometric systems
is larger then the Unimodal systems still it is used in
place where security is the chief concern. By using
appropriate normalization technique and fusion
technique we can achieve a high security multimodal
biometric system .
26. Reference
• M. Indovina, U. Uludag, R. Snelick, A. Mink, and A. Jain,
“Multimodal Biometric Authentication Methods”, Proc.
MMUA 2009, Workshop on Multimodal User Authentication,
pp. 99-106, Santa Barbara, CA, Dec. 11-12, 2009.
• A. Ross and A.K. Jain, “Information Fusion in Biometrics”,
Pattern Recognition Letters, vol. 24, no. 13, pp. 2115-2125,
2003.
• R. Auckenthaler, M. Carey, and H. Lloyd-Thomas, “Score
Normalization for Text-Independent Speaker Verification
Systems”, Digital Signal Processing, vol. 10, pp. 42-54, 2000.