Pattern recognition on line signature seminar Mustansiriya University Department of Education Computer Science

1. ON-LINE SIGNATURE VERIFICATION
Lecture by:
Marwa

2. OUTLINES
1.Introduction
2.Kind of imposter
3.Applications
4.Commercial systems
5.Feature Extraction
6.Enrollment
7.Similarity computation

3. SIGNATURE VERIFICATION
Automatic signature verification is an important
research area because of the social and legal
acceptance and widespread use of the handwritten
signature.
Another advantage of the handwritten signature as
a biometric modality is that it is easily acquired
either with an inking pen over a sheet of paper or by
electronic means with a number of existing pointer-
based devices (e.g., pen tablets, PDAs, Tablet PCs,
touch screens, etc.).

4. SIGNATURE VERIFICATION
Signature verification
 Behavioral biometrics
 Verify user signatures using computers or embedded
devices
 Efficient and effective method of replacing insecure
passwords, PIN numbers, keycards and ID cards
 Advantages
 Customary way of identity verification
 Even advanced PDAs focus pen-input
 People are willing to accept a signature based verification
 Easier, faster, low memory
 Disadvantages
 Dynamic Biometric, Non-repudiation
 Can be forged easily

5. ON-LINE SIGNATURE VERIFICATION
it can be operate in two different ways :
1.Static : in this mode , users write their signature on
paper , digitize it through an optical scanner or a
camera , and the biometric system recognize the
signature its shap . This known as “off- line”.(
geometric features shape)
2.Dynamic: in this mode , user write their signature in
a digitizing tablet which acquires the signature in real
time , dynamic recognition is known as “on-
line“.(geometric features and dynamic features).

6. DIFFERENCE
 Static/Offline
 Early 1970’s
 Only image of signature
 No need of special
hardware, ubiquitous use
 Large storage
 Can not trace speed, style,
pressure etc
 Easier to forge
 Around 95% accuracy
• Dynamic/Online
• Early 1990’s
• Uses shape, speed,
pressure
• Needs special digital
surface, pads and pen etc.
• Numeric data, small storage
• Can use speed, pressure,
angle of pen etc to further
exploit individuality
• Harder to forge
• Around 99% accuracy

7. KIND OF IMPOSTORS
two kinds of impostors are usually considered in
signature verification ;
1. casual impostors (producing random forgeries)
when no information about the target signature is
known.
2.real impostors (producing skilled forgeries)
when some information regarding the signature
being forged is used.

8. APPLICATIONS
The most important applications of on-line signature
biometrics are :
1 .legal (document authentication)
2. medical (record protection)
3 .Banking sectors (cheque and credit card processing).

9. THE MAIN APPLICATIONS
1 .Signature forensics: This is the oldest application of the
handwritten signature , commonly applied to the off-line
image of the written signature..
2. Signature authentication: This type of application
includes system login based on signature, document
encryption, web access.
3. Signature surveillance: The automatic comparison of on-
line signatures can be used to track and detect signers (e.g.,
blacklists of individuals).
4.Digital Rights Management: based on signature.
5.Biometric cryptosystems based on signature: New
developments have demonstrated the feasibility of generating
cryptographic keys based on the time functions of the on-line
signatures

10. COMMERCIAL SYSTEMS
The signature modality is the second behavioral trait
in commercial importance after voice biometric , the
market for signature systems is growing at a faster
rate due to the advent of touch screen portable
devices .
The architecture of on-line signature verification

11. FEATURE EXTRACTION
The existing methods can broadly be divided into two
classes:
1. feature-based: in which a holistic vector
representation consisting of a set of global features is
derived from the signature trajectories, e.g , signature
length ,hight
2 .function-based: in which time sequences
describing local properties of the signature are used
for recognition ,e.g., position trajectory, velocity,
acceleration, force, or pressure

12. ENROLLMENT
Depending on the matching strategy,enrollment can be
divided into two classes:
1 . reference-based :
the features extracted from the set of training signatures
stored as a set of template signatures
each one in the template set corresponding to one
training signature.
The matching process is then performed by comparing the
input signature to each one of the reference Templates
then combining the resulting matching scores with a score-
level fusion technique.

13. ENROLLMENT
2.model-based:
the set of training signatures of a given subject is used
to estimate a statistical model which describes the
behavior of that particular signer.
As in the feature extraction process, the model
complexity can also be adjusted to be user-dependent

14. SIMILARITY COMPUTATION
 Pre-Alignment
is a stage between the input signature and the enrolled
template/model.
In the case of reference-based enrollment, the pre-
alignment is usually conducted before feature extraction
based only on the signature shape.
Techniques following this approach include basic position
and rotation alignment, or more sophisticated
approaches based on boundary warping.
In the case of model-based enrollment, the prealignment
usually consists in the application of a common reference
system.
When no- pre alignment is used , the alignment is either
embedded in the matching procedure or a fixed frame is
used during acquisition in order to have pre-aligned
signatures.

15. Fig. 10.3. Example of local elastic matching of signatures based on DTW (left,
Chinese signature from SVC 2004) and regional modeling based on HMM
(right).

16. SIMILARITY COMPUTATION
 Matching :
In feature-based approaches with reference-based
enrollment, the matching scores are usually obtained by
using some kind of distance measure between the feature
vectors of input and template
signatures or traind classifiers ,distance measure used (e.g
Euclidean distance , weighted Euclidean distance ).
Trained classifiers include approaches like neural network
Function-based approaches can be classified into
local and regional Depending on the matching strategy .

17. SIGNATURE VERIFICATION
 Advantage
1.Social
2.legal acceptance
3.widespread use
4.easily acquired
5.Little time of verification
6.Cheap technology

Disadvantage
1.low universality : as not everyone may be able to sign .
2.low permanence : the handwritten signature tends to vary
along time .
3 .vulnerability to direct attacks using forgeries.
4.There is much precedence for using signature to authentication
document and not for security application .