Offline Signature Verification Based on Fusion of Grid and Global Features Using Neural Networks
ABSTRACT Signature is widely used and developed area of research for personal verification and authentication. In this paper Offline Signature Verification Based on Fusion of Grid and Global Features Using Neural Networks (SVFGNN) is presented. The global and grid features are fused to generate set of features for the verification of signature. The test signature is compared with data base signatures based on the set of features and match/non match of signatures is decided with the help of Neural Network. The performance analysis is conducted on random, unskilled and skilled signature forgeries along with genuine signatures. It is observed that FAR and FRR results are improved in the proposed method compared to the existing algorithm.
 Citations (24)
 Cited In (0)

Conference Paper: Comparative Study of Feature Extraction Methods with KNN for OffLine Signature Verification
[Show abstract] [Hide abstract]
ABSTRACT: In this paper verification of signatures is reviewed and different feature extraction methods with KNN are compared in order to obtain the optimized high performance signature verification for improving the identification rate. The task of signature verification is to judge whether an input signature is a genuine or a forged. This task is performed by comparing the collected signature samples with input signatures. In this purpose, three feature extraction methods are reviewed and used for the comparison of offline signaturesEmerging Technologies, 2006. ICET '06. International Conference on; 12/2006 
Conference Paper: SoundBased Multimodal Person Identification from Signature and Voice
[Show abstract] [Hide abstract]
ABSTRACT: Person identification as a security means has a variety of important applications. Many techniques and automated systems have been developed over the past few decades; each has its own advantages and limitations. There are often tradeoffs amongst reliability, the ease of use, ethical/human rights issues, and acceptability in a particular application. Multimodal identification and authentication can, to some extent, alleviate the dilemmas and improve the overall performance. This paper proposes a new method of the combined use of signatures and utterances of pronounced names to identify or authenticate persons. Unlike typical signature verification methods, the dynamic features of signatures are captured as sound in this paper. The multimodal approach shows increased reliability, providing a relatively simple and potentially useful method for person identification and authentication.Internet Monitoring and Protection (ICIMP), 2010 Fifth International Conference on; 06/2010 
Conference Paper: OffLine Chinese Signature Verification: Using Weighting Factor on Similarity Computation
[Show abstract] [Hide abstract]
ABSTRACT: In this paper we present a new method for offline Chinese signature verification. The approach is based on feature extraction of every segment segmented from the signature image. A signature image is segmented to some segments. Every segment is represented by a set of seven features, because each type feature has a different impact for the signature verification, each type weighting factor is very import for the Similarity Computation of the offline Chinese signature verification. Experimental results show a appropriate weight of the features is promising to distinguish skilled forgeries from genuine signatures effectively.eBusiness and Information System Security (EBISS), 2010 2nd International Conference on; 06/2010
Page 1
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
Offline Signature Verification Based on
Fusion of Grid and Global Features Using
Neural Networks
SHASHI KUMAR D R1, K B RAJA2, R. K CHHOTARAY3, SABYASACHI PATTANAIK4
1Department of CSE, Cambridge Institute of Technology, Bangalore
2Department of ECE, University Visvesvaraya College of Engineering, Bangalore University, Bangalore
3Department of CSE, Seemanta Engineering College, Mayurbhanj, Orissa
4Department of Computer Science, F.M. University, Balasore, Orissa
Abstract:
Signature is widely used and developed area of research for personal verification and authentication. In this
paper Offline Signature Verification Based on Fusion of Grid and Global Features Using Neural Networks
(SVFGNN) is presented. The global and grid features are fused to generate set of features for the verification of
signature. The test signature is compared with data base signatures based on the set of features and match/non
match of signatures is decided with the help of Neural Network. The performance analysis is conducted on
random, unskilled and skilled signature forgeries along with genuine signatures. It is observed that FAR and
FRR results are improved in the proposed method compared to the existing algorithm.
Keywords: Signature, Neural Network, FAR, FRR, Grid, Global, Feature Extraction.
1. Introduction
Biometrics is the science of automatic recognition of individual depending on their physiological and behavioral
attributes. The expansion of networked society and increased use of some personal portable devices like tablet
PCs, PDAs, mobile phones and authorization of access to sensitive data, is demanding the most reliable personal
identification and authentication systems. Among the different forms of biometric recognition systems such as
fingerprint, iris, face, voice, palm etc., signature will be most widely used. The applications like government and
legal financial transaction, bank cheques use signature as one of the personal identification system. The financial
transactions and shopping using debit cards and credit cards require a bill to be confirmed by handwritten
signature. But this leads to increased risk of financial loss due to attempted forgeries. This problem may be
resolved by introducing automatic recognition systems which are being successfully used effectively to analyse
large quantities of biometric data.
Since olden days handwritten signature has been most widely used and accepted individual attributes for
recognition. The design and development of signature recognition system is really big challenge because of the
increased dependence of personal identification systems. Signature recognition system is divided into Online
or dynamic and offline or static recognition. Online recognition refers to a process where the signer uses a
special pen called stylus to create his or her signature, producing the pen locations, speed and pressure, where
as offline recognition deals with signature images acquired by a scanner or a digital camera. In general, offline
signature recognition is a challenging problem, unlike the online signature where dynamic aspects of the
signing action are captured directly as the handwriting trajectory.
Contribution: In this paper, the grid and global features of signature are fused to generate final feature vector
of signature. The Neural Network (NN) is used as a classifier for the verification of signatures.
Organization: The paper is organized into the following sections. Section 2 is an overview of related work.
The SVFGNN model is described in Section 3. Section 4 is the algorithm for SVFGNN system. Performance
analysis of the system is presented in Section 5 and Conclusions are contained in Section 6.
2. Related Work
Javed Ahmed Mahar et al.,[1] used feature extraction methods with KNearest Neighbor for signature
verification. Grid, Global, and Texture Feature Comparison are the three kinds of features used for signature
verification. In the gridbased features, a signature image is divided into rectangular regions and ink distribution
in each region is evaluated. In the global feature comparison, a number of features extracted globally from the
7035
Page 2
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
whole signature are compared. The texturebased features comparison is based on the co occurrence matrices of
the signature image. The Euclidean distance is used for offline signature verification. Taylan Das and Canan
Dulger [2] presented a technique for offline signature verification based on a Neural Network approach trained
with Particle Swarm Optimization algorithm. The three types of fingerprint forgeries such as Random,
Unskilled and Skilled are considered. A trained Neural Network is treated as a specialist in the category of
information.
Ahmad et al., [3] presented an automatic off line signature verification system which uses several statistical
techniques. The algorithm involves building reference model for each local feature extracted from a set of
signature samples during learning phase. In the verification phase two score analysis and normalization
functions were used for fixing the boundary of acceptance and rejection. Li F F [4] proposed the combined use
of signatures and utterances of pronounced names to identify or authenticate individuals. Ji Jun wen et al., [5]
presented the method based on feature extraction from every segment segmented from the signature image.
Every segment is represented by a set of seven features, which has weighing factors. The appropriate weight of
the feature determines whether the signature is skilled forgery or not. Boyko A and Rozorynov G [6] used
signature based authentication. The local global features of the signatures were used to train the Neural
Network. Classification was done by Multi layer Neural Network trained using back propagation with
momentum and bias. Liang Wan et al., [7] described an offline signature verification system that only requires
the genuine signatures of a new user incorporating prior model. At the training stage the system learns the
mapping between the parameters of classifiers without simple forgeries and those with simple forgeries. In the
application stage, a primary classifier is trained for a new user without his/her simple forgeries. The final
classifier is obtained by transforming the primary classifier, via the mapping learnt in the training stage.
Jalal Mahmud and Chowdhury Mofizur Rahman [8] aimed to verify offline signatures using improved
feature analysis and artificial neural network. Feature analyzer can reduce the large domain of feature space and
extract invariable information. For feature extraction, quad tree representation was incorporated with density
analysis, Moment analysis and Structural analysis. For verification from extracted features, multiple feed
forward neural networks are used which are trained in the form of ensemble. Bence Kovari et al., [9] described
an approach for offline signature verification using the enhanced method of Off line signature verification
based on Feature Matching. The method is able to preserve and take an advantage of semantic information
during signature comparison. Feature extraction is characterized by the end points defined by the direction of
their corresponding strokes. The matching of signature is performed by Dynamic Time Warping algorithm and
Mutual Information algorithm. Banshider Majti et al., [10] used the method of geometric centre for feature
extraction of signature. The sub parts of the image are split into vertical splitting and horizontal splitting. Then
24 feature points are extracted for each vertical and horizontal splitting. Euclidean distance is used for
classification. Threshold selection is based on average and standard deviation. Cross validation principle is used
to select reference signatures. Minf Yeng et al., [11] presented a feature selection method based on Contourlet
which is used to capture the structural feature of the signature like directionality and anisotropy information.
The Grid is used to get the the contourlet grid gray property, which gives the statistical feature of signature..
Both the feature vectors are given as input to Support Vectors Machine for training and testing.
I.A.Ismail et al., [12] presented a method for Offline recognition and verification of signatures using
Principal components analysis. The method consists of image prepossessing, feature extraction. The Principal
components analysis is evaluated for the extracted feature. The K nearestneighbors are used in the recognition
process and Neural Network classifier is used in the verification process. Debasish Jena et al., [13] described the
scheme based on selecting 60 feature points from the geometric centre of the signature. The parameters like
mean and variance are used to classify the feature points. These points are compared with trained feature points
for signature verification. Bai Ling Zhang [14] proposed a Kernel Principal Component Selfregression model
for offline signature verification and recognition problems. Developed from the Kernel Principal Component
Regression, the selfregression model selects a subset of the principal components from the kernel space for
verification and recognition. The model directly works on bitmap images. Each user is assigned an independent
Kernel Principal Component Selfregression model for coding the corresponding visual information.
Vu Nguyon et al., [15] proposed an effective method to perform offline signature verification based on
intelligent techniques of Neural Classifiers and Support Vector Machines. Structural features are extracted from
the signatures Contour using the Modified Direction Feature and its extended version: the Enhanced Modified
Direction Feature. The neural networkbased techniques and Support Vector Machines were investigated and
compared for the purpose of signature verification. Stephane Armand et al., [16] presented an effective method
to perform offline signature Verification and identification. The signature's contour is first determined from its
7036
Page 3
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
binary representation. Unique structural features are subsequently extracted from the signature's contour through
the use of a combination of the Modified Direction Feature. The other features like Centroid, Trisurface and
Length are also extracted. To classify the signatures, the Resilient Back Propogation Neural Network and Radio
Basic Function Network are used. Danjun pu and Sargur N Srihari [17] presented a prabobalisitc measure for
signature verification using Bayesian learning process. Features such as gradient, structural and concavity were
used and the distance between two signatures was computed by the correlation similarity measurement. The
distribution of the genuine and forgery signature were taken as Gausian. The Bayesian Learning algorithm was
used to learn the prior distributions of the signatures. The probabilities of the query and log likelihood ratio of
the query was determined for both genuine and forgery classes. Luana Batista et al., [18] proposed a two stage
offline signature verification system based on dissimilarity representation. In the first stage a discrete set of left
to right Hidden Markov Models representing both genuine and forged classes of different number of states and
different code book sizes were used to measure similarity values which form feature vectors. In the second stage
these vectors were used to train a classifier to find the decision between genuine and forgery. Michal Papaj and
Ewa Hermanowicz [19] used cross correlation approach and dynamic time wrapping for handwritten signature
verification. In the algorithm the signature was treated as a complex trajectory and used on xy coordinates.
Bhupendra M Chaudhari et al.,[20] proposed signature verification system using Fuzzy Min – Max Neural
Network. The input signature image is preprocessed and normalized using Hu’s seven moment invariant
methods to make them invariant to position, translation, rotation and shear. The normalized image is applied to
fuzzy min – max neural network for classification.
Muhammed Reza Pourshahabi et al., [21] presented offline handwritten signature identification and
verification using contourlet transform. Noise removal was performed in the preprocessing stages of image
enhancement. After size normalization, contourlet transform was applied to compute the contourlet coefficients
and feature vector was formed. Classification was performed using the Euclidean distance classifier. Sepideh
Afsardoost et al., [22] introduced the offline signature verification system based on geometric centroid features.
The data base signatures and test signature were preprocessed for noise removal. The signatures are then
subjected vertical, horizontal and diagonal splitting around the geometric centers. Euclidean distance model was
used for classification. Sharifah Mumtazah Syed Ahmad et al., [23] presented an automatic off – line signature
verification system using different statistical techniques. HMM based modeling was used to build a reference
model for each of the local features extracted from signature image. The three layers of statistical techniques
were used in the verification. The first stage involves the HMM based log likelihood probability matching score.
The second stage maps these scores into soft boundary ranges of acceptance or rejection using zscore analysis.
Classification was done by Bayesian inference technique.
3. Proposed Model
The block diagram of Offline Signature Verification Based on Fusion of Grid and Global Features Using
Neural Networks (SVFGNN) discussed in detail and is as shown in the Figure 1. It is divided into three phases
viz., Preprocessing Phase, Feature Extraction Phase, and the Verification Phase.
3.1 Test Signature
It is the signature of a person whose authenticity is to be verified.
3.2 Database
The signatures are collected using either black or blue ink, on a white A4 sheet of a paper, with four signatures
per page. Using the scanner the four signatures are digitized, with 96dpi resolution in 256 grey levels. A
group of 20 persons are used to collect 30 specimens from each persons resulting in 600 signature samples.. The
genuine signatures are collected from 10 persons and the forged signatures are collected from the remaining 10
persons.
7037
Page 4
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
Figure 1: Block Diagram SVFGNN
3.3 Preprocessing
The features of signature are extracted using preprocessing stages such as (i) Noise Reduction (ii) Size
Normalization, and (iii) Skelitonization.
3.3.1 Noise reduction
Imperfection in the scanner intensity of light, scratches or dirt on the camera or scanner lens etc., introduces
noises in the scanned signature images. A filtering function is used to remove the noises in the image. It is
required to eliminate single white pixels on black background and single black pixels on white back ground. In
order to eliminate the noise we apply a 3 x 3 mask to the image with a simple decision rule: if the number of the
8neighbors of a pixel that have the same color with the central pixel is less than two, then reverse the color of
the central Pixel. The Gaussian filter is used for the noise removal. Since Gaussian function is symmetric,
smoothing is performed equally in all directions, and the edges in an image will not be biased in particular
direction. The signature before and after removal of noise are as shown in the Figure 2 and 3 respectively.
Figure 2. signature before noise reduction.
Figure 3. signature after noise reduction
Authentication
Preprocessing
Feature Extraction
Test and Database
signatures
NN Training
Verification
ISSN: 097554627038
Page 5
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
3.3.2 Size normalization
Normally any person while putting his signature uses an arbitrary baseline. The positional information of the
signature is normalized by calculating an angle θ about the centroid (x,y) such that rotating the signature by θ
brings it back to a uniform baseline. The size normalization in offline signature verification is important because
it establishes a common ground for image comparison. Taylor’s maximization is used for normalization.
The mean µx of the x –series pixels is calculated by Equation 1.
??? ∑ ??
Where t varies from t = 1 to t = y .
The function f(θ) is defined as shown in the Equation 2
?
?
?
? .….(1)
???? ? ?? ???? ?? ??
?
…?2?
Where xt
is performed using the Equation 3.
??
The function f (θ ) is defined by substituting Equation 3 in Equation 2.
?
* indicates a rotated x value. To preserve the centroid, rotation about the coordinate (µx , µy)
?? ???? ??? cos ? ? ???? ??? sin? ? ?? ...... (3)
???? ? ?????? ???cos? ? ???? ???sin? ? ??? ?? ??
?
?
? 2???? cos?sin? ? ???sin??
? ???
? cos??
?
? cos?θ???
?
?
?
? 2cos?sin??????
?
?
? ?????????
?
?
Where ?? ? ?? ? ??
and ?? ? ?? ? ??
? ? ???? ? ????
?
?
? ? ?? ?? ? ????
?
?
? ? ???? ? ???
?
?
???? ???
The derivative of the Equation 4 gives Equation 5
f??θ? ? 2Qcos?θ ? 2Pcosθsinθ ? 2Rcosθsinθ ? 2Qsin?θ ….(5)
The roots of Equation 5 are given in Equation 6 and 7
???? 1 ? ? ???????
?
√??
???????
??? ?? ?? ???? ??? ……...(6)
???? 2 ? ? ???????
?
√??
???????
??? ?? ?? ???? ??? …… (7)
7039
Page 6
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
Obtaining the smallest value of the roots ie., θ from Equation 6 and 7 results in maximum value of f (θ).
Further refinements make the time series to evolve in a consistent direction so that size normalization is
perfectly achieved. Figure 4 and 5 shows the signature samples before normalization and after normalization
respectively.
Figure 4. Signature before normalization
Figure 5. Signature after the normalization
3.3.3 Skeletonization
Reducing image to its single pixel width is called as skeletonization. The generation of a skeleton is realized by
applying an iterative process which erodes the object layer by layer until only the object spines, which form the
skeleton remains, this iterative process is called thinning. A thinning algorithm contains a set of pixel deleting
conditions, which enable it to erode the object iteratively. The skeletonization steps are as follows.
Step 1: mark all the points of the signature that are candidates for removing (black pixels that have at least one
black with 8white neighbor and at least two black with 8white neighbors).
Step 2: Examine one by one pixel following the contour lines of the signature image, and remove them as their
removal will not cause a break in the resulting pattern.
Step 3: If at least one point was deleted then go back to Step 1 and repeat the process once again.
Skeletonization makes the extracted features invariant to image characteristics like the qualities of the pen, the
paper, the signer, the digitizing method and quality. The Skeletonization process supports the following
properties, so that the thinning result can be characterized as a skeleton of the 2D binary object.
Geometry preservation is a major concern of thinning algorithms. For example, an object like b should not
be converted into an object like o. To preserve the geometry of the original image, a thinning algorithm must
contain certain geometry preserving conditions.
Topology preservation is the second major concern of thinning algorithms. For example, an object like o
should not be converted into an object like c.
Finally, a skeleton of an object should be, ideally, as thin as possible i.e., one pixel wide and represent the
object through its spine or medial axis.
The Figures 6 and 7 shows the signature sample before skelitonization and after skelitonization process
respectively.
Figure 6. Signature before skeletonization
ISSN: 097554627040
Page 7
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
Figure 7. Signature after skeletonization
3.4 The Feature Extraction Phase
The choice of a powerful set of features is crucial in signature verification systems. The features used must be
suitable for the application and for the applied classifier. In this system, two groups of features are used such as
grid features and global features. For grid information features, the image is segmented into appropriate number
of rectangular regions. The global features provide information about specific cases concerning the structure of
the signature.
3.4.1 Grid Feature
Grid segmentation procedures have been used extensively in the offline signature verification approach. The
skeletonized image is divided into 120 rectangular segments (15x8), and for each segment, the area (the sum of
foreground pixels) is calculated. The results are normalized so that the lowest value ie., the rectangle with the
smallest number of black pixels would be zero and the highest value i.e., the rectangle with the highest number
of black pixels would be one. The resulting 96 values form the grid feature vector.
It is very encouraging to recognize diagonally so that more points may be diagnosed for generating the
vector matrix to get results more accurate than the simple grid.
Figure 8. Simple Grid with Image
3.4.2 Global Feature:
Some common global features discussed below are implemented in our experiments.
Aspect Ratio: The ratio of signature pure height to signature pure width.
Signature height: It is the height of the signature image, after width normalization.
Image area: It is the number of black pixels in the image. In skeletonized signature images, image area
represents a measure of the density of the signature traces.
Pure width: The width of the image with horizontal blank spaces removed.
Pure height: The height of the signature image after vertical blank spaces removed.
3.5 Neural Network (NN) Training:
The standard back propagation neural network classifier for verification is used. Standard back propagation is a
gradient descent algorithm, as is the WidrowHoff learning rule, in which the network weights are moved
along the negative of the gradient of the performance function. The term back propagation refers to the
manner in which the gradient is computed for nonlinear multilayer networks. Typically, a new input leads to an
output similar to the correct output for input vectors used in training which are same as the new input being
presented. This generalization property makes it possible to train a network on a representative set of
input/target pairs and get good results without training the network on all possible input/output pairs.
3.6 Verification process
Multilayer feed forward artificial neural network for verification of offline digitized signatures is used. The
proposed NN consists of 30 input variables which are extracted from signature features, and it is designed to
verify one signature at a time. Back propagation algorithm is used for training.
4. Algorithm
Problem Definition: Given a test signature for the verification of authenticity. The objectives are:
7041
Page 8
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
(i) The database signatures and test signature are preprocessed to eliminate noise.
(ii) The database signatures are considered and the neural network is trained by the feature of each signature.
(iii) The test signature features are compared with the database signature features using NN.
Assumptions:
(i) The grid size of 15*8 is considered.
(ii) The signature image of type 96dpi resolution in 256 gray levels.
The algorithm of SVFGNN is explained in the Table 1 for the signature verification using Grid and Global
features by training NN.
5. Performance Analysis
The test and the database signatures of 96dpi resolution in 256 grey levels are considered for the performance
analysis. It is observed from the Table 2. that the values of FRR (False Rejection Rate) and FAR (False
Acceptance Rate) are improved in the proposed method of combined offline signature verification using neural
Table 1: Algorithm of SVFGNN
network compared to existing grid and global feature method. The FAR is improved by a factor 2530%
compared to grid Feature using KNN Classification [1] and 5060% as compared to Global Feature using KNN
Classification method. The FRR is improved by a factor 712% in the proposed algorithm compared to grid
Feature using KNN Classification and 3035% as compared to Global Feature using KNN Classification
method.
Table 2: FAR and FRR for different algorithms.
Method
Grid Feature using
KNN Classification
Global Feature using
KNN Classification
Proposed SVFGNN
algorithm
FRR
FAR
8.07%
5.91%
11.26%
9.53%
7.51%
4.16%
6. Conclusion
Signature is a behavioral biometric used to authenticate a person in day to day life. We proposed the algorithm
which fuses both the global and grid features to yield better results as compared to individual global features
and grid features. The Back Propagation Neural Network is used for verification of offline signatures. It is
observed that the values of FRR and FAR are improved in the proposed algorithm compared to the existing
algorithm. In future signature may be converted into transform domain for the verification of the performance
analysis.
Input: Reference Signature s and Test Signature
Output: Verified Signature
Step 1: Noise Reduction, Normalization and Skelitonization is performed on database
signature and test signature.
Step 2: Global Features and Grid Features of database signatures are extracted.
Step 3: Neural Network is trained by features of database signatures.
Step 4: Global Features and Grid Features of test signature are extracted.
Step 5: Compare features of test signature with the features of database signatures using NN.
Step 6: Authentication of test signature.
7042
Page 9
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
References
[1] Javed Ahmed Mahar, Mumtaz Hussain Mahar, Muammed Khalid Khan. (2006): Comparative Study of Feature Extraction for
OffLine Signature Verification, Second International Conference on Emerging Technologies, pp. 115120.
[2] Taylan Das, M., L Canan Dulger, L. (2007): Off Line Signature Verification with PS0NN Algorithm, Twenty Second
International Symposium on Computer and Information Sciences, vol 7, issue 9, pp. 16.
[3] Ahmad, S. M. S., Shakil, A., Faudzi, M. A., Anwar, R. M., Balbed, M. A. M. (2009): A Hybrid Statistical Modelling,
Normalization and Interferencing Technuques of Offline Signature Verification System, World Congress on Computer Science
and Information Engineering, vol 6, pp. 611.
[4] Li, F. F. (2010): Sound Based Multimodal Person Identification from Signature and Voice, Fifth IEEE international Conference
on Internet Monitoring and Protection, pp. 8488.
[5] Ji Jumwen, Chen Chunbo, Chen xiansu. (2010): Offline Chinese Signature Verification: Using Weighting Factor on Similarity
Computation, Second IEEE International Conference on eBusiness and Information System security, pp. 14.
[6] Boyko, A., Rozorynov, G. (2010): Signature based Authentication, IEEE International conference on Modern Problems of Radio
Engineering, Telecommunications and Compute Science, pp. 300.
[7] Liang Wan, Zhouchen Lin, Rongchun Zhao. (2003): Off Line Signature Verification Incorporating the Prior Model,
Proceedings of the Second International Conference on Machine Learning and Cybernatics, pp. 16021606.
[8] Jalal Mahmud, Chowdhury Mofizur Rahman. (2005): On Analysis of Multi Dimensional Features for Signature Verification,
International Conference on Computational Intelligence for Modelling, Control and Automation and International Conference on
Intelligent Agents, Web Technologies and Internet Commerce, pp. 735740.
[9] Bence Kovari, Zsolt Kertesz, Attila Major. (2007): Off Line Signature Verification Based on Feature Matching, Eleventh
International Conference on Intelligent Engineering System, pp. 93 – 97.
[10] Banshider Majhi, Santhosh Reddy, Y., Prasanna Babu, D. (2006): Novel Features for Offline Signature Verification,
International Journal of Computer Communications and Controls, vol . 1, no. 1, pp. 17 – 24.
[11] Ming Yang, Zhongke Yin, Zhi Zhong, Sheng Shu Wang, Pei Chen Yang Sheng Xu. (2002): A Contourlet Based Method for
Handwritten Signature Verification, National Conference on RealTime Imaging, pp. 227236.
[12] Ismail, I. A., El danf, T., Ramadan, M. A., Samak, A. H. (2008): Automatic Signature Recognition and Verification using
Principal Component Analysis, Fifth International Conference on Computer Graphics, Imaging and Visualization, pp. 356361.
[13] Debasish Jena, Banshidhar Majhi, Sanjay Kumar Jena. (2008): Improved Offline Signature Verification Scheme Using Feature
Point Extraction Method, Journal of Computer Science, vol. 4, issue 2, pp. 111 – 116.
[14] Bai ling Zhang. (2006): Off Line Signature Recognition and Verification by Kernel Principal Component Self Regression, Fifth
International Conference on Machine Learning and Application, pp. 2833.
[15] Vu Nguyen, Michael Blumenstein, Vallipuram Muthukkumarasamy, Graham Leedham. (2007): Off Line Signature Verification
using Enhanced Modified Direction Features in conjuction with Neural Classifiers and Support Vector Machine, Ninth
International Conference on Document Analysis and Recognition , vol 2, pp.734738.
[16] Stephane Armand, Michael Blumanstein, Muthukkumarasamy, V. (2006): Off Line Signature Verification based on Modified
Direction Feature, Eighteenth International conference on Pattern Recognition, vol 4, pp. 509512.
[17] Danjun Pu, Sargur Srihari, N. (2010): A Probabolisitic Measure for Signature Verification based ob Bayesian Learning, IEEE
International Conference on Pattern Recognition, pp. 118891.
[18] Luana Batista, Eric Granger, Robert Sabourin. (2010): Applying Dissimilarity Representation to Off line Signature Verification,
IEEE International Conference on Pattern Recognition, pp. 1293 – 1298.
[19] Michal Papaj, Ewa Hermanowicz. (2010): Identity Verification Using Complex Representations of the Handwritten Signatures,
Second IEEE International Conference on Information Technology, pp. 79 – 82.
[20] Bhupendra Chaudhari, M., Atul Barhate, A., Anita Bhole, A. (2009): Signature Recognition Using Fuzzy Min – Max Neural
Network, IEEE International Conference on Control Automation, Communication and Energy Conservation, pp. 17.
[21] Muhammad Reza Pourshahabi, Mohamad Hoseyn Sigari, Pourreza, H. R. (2009): Offline Handwritten Signature Identification
and Verification Using Contourlet Transform, IEEE International Conference on Soft Computing and Pattern Recognition, pp.
670 – 673.
[22] Sepideh Afsardoost, Siamak Yousefi, Mohammed Ali Khorshidi. (2008): Offline Signature Verification Using Geometric Center
Features, IEEE International Conference on Signal Processing, pp. 1491 – 1494.
[23] Sharifah Mumtazah Syed Ahmad, Asma Shakil Masyura Ahmad Faudzi, Anwar, R. M., Muhamad Balbed,M. A. (2008): A
Hybrid Stastical Modelling Normalization and Interferencing Techniques of an Offline Signature Verification System, World
Congress on Computer Science and Information Engineering, pp. 6 – 11.
Shashikumar D R received BE degree in Electronics & Communication Engineering from
Mysore University and ME degree in Electronics from Bangalore University, Bangalore. He is
pursuing his Ph.D. in Information and Communication Technology of Fakir Mohan University,
Balasore, Orissa under the guidance of Dr. K. B. Raja, Assistant Professor, Department of
Electronics and Communication Engineering, University Visvesvaraya College of Engineering,
Dr.Sabyasachi Pattanaik Reader & HOD, Department of Information and Communication
Technology F M University, Balasore, Orissa R K Chhotaray, Principal, Seemantha Engineering College,
Orissa. He is currently working as Professor, Dept. of Computer Science, Cambridge Institute of Technology,
and Bangalore. His research interests include Microprocessors, Pattern Recognition, and Biometrics.
K B Raja is an Assistant Professor, Dept. of Electronics and Communication Engineering,
University Visvesvaraya college of Engineering, Bangalore University, Bangalore. He obtained
his BE and ME in Electronics and Communication Engineering from University Visvesvaraya
College of Engineering, Bangalore. He was awarded Ph.D. in Computer Science and
Engineering from Bangalore University. He has over 45 research publications in refereed
International Journals and Conference Proceedings. His research interests include Image
7043
Page 10
Shashi Kumar et. al. / International Journal of Engineering Science and Technology
Vol. 2(12), 2010, 70357044
ISSN: 09755462
Processing, Biometrics, VLSI design, Signal Processing and computer networks.
Dr. Sabyasachi Pattnaik has done his B.E in Computer Science, M Tech., from IIT Delhi. He
has received his Ph D degree in Computer Science in the year 2003 & now working as Reader
in the Department of Information and Communication Technology, in Fakir Mohan University,
Vyasavihar, Balasore, and Orissa, India. He has got 20 years of teaching and research
experience in the field of neural networks, soft computing techniques. He has got 42
publications in national & international journals and conferences. He has published three books in office
automation, object oriented programming using C++ and artificial intelligence. At present he is involved in
guiding 8 Ph D scholars in the field of neural networks, cluster analysis, bioinformatics, computer vision &
stock market applications. He has received the best paper award & gold medal from Orissa Engineering
congress in 1992 and institution of Engineers in 2009.
R K Chhotaray received B.Sc Engineering in Electrical Engineering and M.Sc Engineering in
Electrical Engineering with specialization in Control Systems from Banaras Hindu University,
and Ph D in Control Systems from Sambalpur University. He was Professor and Head of
Department of Computer Science and Engineering, Regional Engineering College, Rourkela,
from which he retired in 2003. Currently he is working as Principal of Seemanta Engineering
College, Orissa. He has been associated with many Universities of India in the capacity of
Chairman and member of various Boards of Studies, syllabus committee, and Regulation committee. He has
about hundred publications in International and National journals of repute, and has received Best Technical
Paper award in many occasions. His special fields of interest include Control of Infinite dimensional Hereditary
Systems, Modeling and Simulation, Theoretical Computer science, signal and Image processing, and
optimization.
7044
View other sources
Hide other sources
 Available from Raja K B · Oct 2, 2014
 Available from ijest.info