Online Signature Verification using PCA and Neural Network - IJRIT

IJRIT International Journal of Research in Information Technology, Volume 1, Issue 5, May 2013, Pg. 65-73

International Journal of Research in Information Technology (IJRIT)

www.ijrit.com

ISSN 2001-5569

Online Signature Verification using PCA and Neural Network 1

1

Jyoti Bhalla

Student, MTech , Department of Information Technology, Guru Gobind Singh Indraprastha University, Delhi 1

[email protected]

Abstract The purpose of this paper is to study about the use of distinctive anatomical biometric characteristic- Signature for automatically recognizing individuals. Biometric techniques have been used for personal identification in the past. Various identification methods include iris, retina, face, fingerprint, and signature-based identification. The non-vision based techniques include face recognition, fingerprint recognition, iris scanning and retina scanning and the vision-based ones include voice recognition and signature verification. Signature has been a distinguishing feature for person identification because of its ease of acceptance by the public, and lower implementation cost. Signature verification can be applied in commercial fields such as E-business, which includes online banking transactions, electronic payments, access control, and so on. In this paper Online signature recognition and verification using Neural Network is proposed. PCA is used to extract features used for training the network. A verification stage includes applying the extracted features of test signature to a trained neural network which will classify it as a genuine or forged.

Keywords: Biometrics, Online signature, PCA, Neural Network.

1. Introduction Biometrics is an emerging field of technology. It makes use of unique but measurable physical, biological, or behavioral characteristics to perform identity verification of a person. A number of biometric techniques have been proposed for personal identification in the past. Well known biometric methods include iris, retina, face, fingerprint, and signature-based identification. Fingerprint recognition, face recognition, iris scanning and retina scanning fall under the non-vision based techniques. Among the vision-based ones, we can mention voice recognition and signature verification.

Jyoti Bhalla, IJRIT

65

Handwritten signature is one of the most widely accepted personal attributes for identity verification of the person. The written signature is regarded as the primary means of identifying the signer of a written document based on the implicit assumption that a person’s normal signature changes slowly and is very difficult to erase, alter or forge without detection. The handwritten signature is one of the ways to authorize transactions and authenticate the human identity compared with other electronic identification methods such as retinal vascular pattern screening and fingerprints scanning. People can easily migrate from using the popular pen-and-paper signature to one where the handwritten signature is captured and verified electronically.

Signature verification is the task of authenticating a person based on his handwritten signature. There have been two types of signature verification systems in the literature Off-line and On-line. a.

On-line signature verification system is dynamic system

Where a signature is written onto an interactive electronic device such as a tablet and is read online, which is then compared to the signatures on file of the person to check for authenticity. Online (dynamic) signature [4, 5, 6, 7, and 8] uses signatures that are captured by pressure-sensitive tablets that extract dynamic properties of a signature in addition to its shape. Many important features are utilized with online signatures that are not available for the offline ones. In this process that the signer uses a special pen called a stylus to create his or her signature, producing the pen locations, speeds and pressures. b.

Offline signature verification system is static system

Where a signature is written offline such as bank checks and the system read the image scan then verifies it with the signatures on file for the customer. Offline (static) signature [1, 2, 3] takes the image as an input of a signature and it is useful in bank check and document. In this process signature images is acquired by a scanner or a digital camera.

In general, off-line signature recognition is a challenging problem. Unlike the on-line signature, where dynamic aspects of the signing action are captured directly as the handwriting trajectory, the dynamic information contained in offline signature is highly degraded. Handwriting features, such as the handwriting order, writing-speed variation, and skillfulness, need to be recovered from the grey level pixels.

1.1 Problem Identification

Handwritten signature is one of the most widely accepted symbol of consent and authorization, especially in the prevalence of credit cards and bank cheques has long been the target of fraudulence. Therefore, with the growing demand for processing of individual identification faster and more accurately, the design of an automatic signature verification system faces a real challenge. Signatures themselves are worthless but because they are used as authorization of various things like cheques and wills, they become targets of forgers.

1.1.1 Types of Forgeries There are three different types of forgeries to take into account. First one is random forgery which is written by the person who doesn’t know the shape of original signature. The second, called simple forgery, which

is represented by a signature sample, written by the person who knows the shape of original signature Jyoti Bhalla, IJRIT

66

without much practice. The last type is skilled forgery, represented by a suitable imitation of the genuine signature model [4]. Each type of forgery requires different types of verification approach [5]. Hybrid systems have also been developed [10]. Figure 1 shows the different types of forgeries and how much they vary from original signature. The type of forgeries that can be encountered in signature are

Forgeries

Random Forgeries

Simple Forgeries

Skilled Forgeries

Hybrid Forgeries

Professional Forgeries

Amateur Forgeries

Figure 1 Different types of forgeries

Figure2. Example Different types of forgeries (A) Original signature, (B) Random forgery, (C) Simple forgery, (D) Skilled forgery

Computer detection of forgeries may be divided into two classes, the on-line approach and the off-line approach. The on-line approach of signature verification involves the use of dynamic information which is attained during the actual signing of the signature. This information includes the velocity of the pen, the pressure exerted during the signing process and the acceleration of the pen. The off-line approach uses only the final result of the signing process, i.e. the image of the signature.

Jyoti Bhalla, IJRIT

67

2. Literature Review Handwriting is very complex in nature. It is a biomechanical process that includes the movements of fingers, wrist, and forearm. It has been shown that humans generate handwriting through controlling the magnitude and direction of speed [11]. Therefore, many techniques and models are proposed and applied for online signature verification system using one type of feature or both. A wide range of methods for on-line handwritten signature verification have been reported, but more work has been done on off-line verification. Depending on the signature capture device used, features such as velocity, pen pressure and pen tilt are used in on-line verification in addition to spatial (derived from (x; y) coordinates) features. Different approaches can be categorized based on the model used for verification. Most of the approaches do a fair amount of preprocessing before extracting features from the signature. The most common method to find the similarity between the input feature vector and the stored template is to use some variant of the Euclidean distance. Since the number of points differs between any two signatures, some form of string matching [4,5] is used. Hidden Markov models, well known for their success in speech recognition, have also been successfully applied to handwriting recognition. For signature verification, a variety of models [6,7] and features [8,9] have been evaluated. The number of signatures captured for a user during the enrolment phase varies between 6 and 20. The equal error rate generally lies between 1% and 6%. Since there does not exist a signature database in the public domain, every research group has collected its own data set, having between 9 and 105 individuals enrolled. This makes a comparison of different signature verification systems a difficult task. Two-class pattern recognition problem for online signature verification has been proposed by [13]. First, they experimented with Bayes classifier on the original data, as well as a linear classifier used in conjunction with Principle Component Analysis (PCA). The Back Propagation Neural Network (BPNN) used by [14] per each signer with parameters as follows: (i) 72 signatures for recognition, (ii) 251 neural Bias in the input layer, (iii) 11 neural Bias in the hidden layer, (iv) 2 neural in the output layer, (v) 60 patterns in train stage. An online signature verification based on Parzen Window Classifier (PWC) and Hiden Markov Model (HMM) have been developed by [15]; the application of HMM to time sequences directly based on the dynamic functions. Reference [16] multiple classifiers (neural network, support vector machine and Pearson correlation) were used to develop an on-line signature verification system. Then, they fused them by applying a different fusion technique. A new technique for online signature verification has been presented by [17]. The technique integrates a longest common subsequences (LCSS) detection algorithm which measures the similarity of signature time series into a kernel function for support vector machines (SVM). 2.1 Signature Verification Steps The major steps involved in signature verification include Data acquisition, Pre-processing, Features Extraction which provide the input for Verification of data.

Data acquisition

Preprocessing

Features Extraction

Verification

Figure 3 Steps involved in signature verification

Jyoti Bhalla, IJRIT

68

In each step after getting data there are many methods and techniques that can be applied individually or by combining two or more techniques to get a better results. The state of the art in signature verification follows a pattern that is similar to image processing

Figure 4: Signature Verification System Work Flow

In the classification phase, personal features extracted from an input signature are compared with template signature stored in the knowledge base, to check the authenticity of the test signature

2.2 Data Acquisition and Pre-processing The online signature recognition, where signatures are acquired during the writing process with a special instrument, such as pen tablet. The input signatures are pre-processed, and then the personal features are extracted and stored into the knowledge base. In fact, there is always dynamic information available in case of online signature recognition, such as velocity, acceleration and pen pressure. So far there have been many widely employed methods developed for online signature recognition for example, Artificial Neural Networks (ANN), dynamic time warping (DTW), the hidden Markov models (HMM).

2.3 Feature Extraction Features can be classified as Global features, Local features or a combination of both. Global features describe properties of the whole signature. Examples of global features include total writing time, bounding box or the number of strokes. (A stroke is the sequence of points through which the pen moves while touching the paper. A signature is usually made up of several strokes.) Local features are properties which refer to a position within the signature. Distance or curvature change between consecutive points on the signature trajectory is also considered as local feature [4]. The use of global features [3] alone has the advantage that the verification time is very short, but the error rates of algorithms that also incorporate local features are generally lower.

Jyoti Bhalla, IJRIT

69

Local features

Global features

Hybrid Features

Feature Extraction Figure 5 Feature Extraction

The off-line recognition just deals with signature images acquired by a scanner or a digital camera. In general, offline signature recognition& verification is a challenging problem. Unlike the on-line signature, where dynamic aspects of the signing action are captured directly as the handwriting trajectory, the dynamic information contained in off-line signature is highly degraded. Handwriting features, such as the handwriting order, writingspeed variation, and skillfulness, need to be recovered from the grey-level pixels.

2.4 Verification In an online or an offline signature system, two steps are required before the final decision is made; first, a user is registered by providing samples of signature (reference signatures), then, when a user presents a signature (test signature) claiming to be a particular individual, the test signature is then compared with reference signatures for that user. If the similarity is very near to a certain threshold, the user is accepted otherwise rejected. On-line data records the motion of the stylus while the signature is produced, and includes location, and possibly velocity, acceleration and pen pressure, as functions of time. Online systems use this information captured during acquisition. These dynamic characteristics are specific to each individual and sufficiently stable as well as repetitive. Off-line data is a 2-D image of the signature. Processing Off-line is complex due to the absence of stable dynamic characteristics. Difficulty also lies in the fact that it is hard to segment signature strokes due to highly stylish and unconventional writing styles. The nonrepetitive nature of variation of the signatures, because of age, illness, geographic location and perhaps to some extent the emotional state of the person, accentuates the problem. All these coupled together cause large intrapersonal variation.

2.5 Performance Signature Verification Method Performance of any signature verification method can be measured using two parameters : FAR and FRR. The FRR is the ratio of the number of genuine test signatures rejected to the total number of genuine test signatures submitted. The FAR is the ratio of the number of forgeries accepted to the total number of forgeries submitted. False Acceptance Rate (FAR) and False Rejection Rate (FRR) are represented by the equations as given:

Jyoti Bhalla, IJRIT

70

FAR = FRR =


    
     
     
      

× 100

× 100

It should have an acceptable trade-off between a low False Acceptance Rate (FAR) and a low False Rejection Rate (FRR). The false rejection rate (FRR) and the false acceptance rate (FAR) are used as quality performance measures. Choosing of a threshold plays a major role as alteration in the threshold to decrease the FRR results in invariably increase in the FAR and vice versa. 2.6 Feature Classification Hybrid Approach Basically, there are well known steps involved in online signature verification as mentioned in section 1 and detailed in [4, 9, 10, 16]. Researchers are focusing on selecting powerful features [4, 12, 13] and a hybrid method [15, 16, 17] to guarantee a higher accuracy of verifying identity of persons via signature. In this paper, two classifiers have been used. First one is used to classify the global features that are chosen; the second one is used to classify the local features. The results of first classifiers are fed as input to the other to get the final decision. 2.6.1 Global Classifier Thirteen global features are selected to be classified by this model which is: Width, Height, Height Width Ratio, Total Area, Max Azimuth, Average Azimuth, Max Altitude, Average Altitude, Max Pressure, Average Pressure, Number of Times Pen Up, Total time of signature, Total distance travelled by the pen. In order to obtain the threshold of the global features classifier for a specific user, the probability score of the reference signatures is calculated. Threshold of a specific user is the minimum values among all reference signatures, and it is kept in his profile. In the test phase, all steps are repeated until getting the Probability Score (PS) value; if the current PS value is greater than or equal to the threshold value stored in the user profile, the user is accepted as genuine, otherwise is rejected as a forgery. First, getting all features values of all reference signatures, then, the mean and variance are calculated. Neural Network is designed using a three-layer MLP, back-propagations to recognize the signatures. 2.6.2 Probabilistic model PCA is applied to extract the important features and train the same network for the selected features. Test signature passes through the four trained forward network only to obtain the test error for comparing with train error. 2.6.3 Combination The results of probabilistic model classifier can be fed to neural network classifier. In order to get the benefit from both classifiers, thus the decision-making can be a combined score of both the models. Each one of the proposed classifier modules is characterized by its own characteristics. A user is considered as a genuine if he passes in both classifiers, otherwise, he is considered as a forgery. The two measures FRR and FAR can be used to evaluate the performance of proposed verification system

Jyoti Bhalla, IJRIT

71

3. Conclusion In this paper, the probability model and BPNN has been proposed. The combination model overcomes the drawbacks of using each model individually. The obtained result can have benefits of both models. Generally, using different dataset yield different result of FRR and FAR even if the same approach is used.

4. References [1] M. K. Kalera, S. Srihari, and A. Xu , “Offline signature verification and identification using distance statistics,” Int. J. Pattern Recognit . Artif.Intell. (IJPRAI), vol. 18, no. 6, pp. 1339–1360, 2004. [2] R. Sabourin , M. Cheriet, and G. Genest, “An extended-shadow-code based approach for offline signature verification,” in Proc. 2nd Int. Conf. Doc. Anal. Recognit. (ICDAR-2), Tsukuba Science City, Japan, Oct.1993, pp. 1–5. [3] J. Pan and S. Lee,” Offline Tracing and Representation of Signatures,” Proc. CVPR, pp. 679–680, 1991. [4] A. K. Jain, F. D. Griess, and S. D. Connell, “On-line signature verification,” Pattern Recognit., vol. 35, no. 12, pp. 2963–2972, Dec. 2002. [5] V. S. Nalwa, “Automatic on-line signature verification,” Proc. IEEE, vol. 85, no. 2, pp. 215–239, Feb. 1997. [6] R. S. Kashi, J. Hu, W. L. Nelson, and W. L. Turin, “A hidden Markov model approach to online handwritten signature verification,” Int. J. Doc. Anal. Recognit. (IJDAR), vol. 1, no. 2, pp. 102–109, 1998. [7] M. E. Munich and P. Perona, “Visual identification by signature tracking,” IEEE Trans. Pattern Anal.Mach. Intell. (T-PAMI), vol. 25, no. 2, pp. 200–217, Feb. 2003. [8] T. Hastie, E. Kishon, M. Clark, and J. Fan, "A model for signature verification", Proc. 1991 IEEE Int. Conf. on Syst., Man, Cybern., vol. 1, pp.191 - 196 , 1991. [9] A. Kholmatov,” Biometric Identity Verification Using on-line & off-line signature Verification,” Master’s thesis, Turkey. Sabanic University, p. 17, 2003. [10] G. V. Kiran, R. S. Kunte, and S. Samuel, “On-line signature verification system using probablistic feature modeling,” in Proc. 6th Int. Symp. Signal Process. Appl., Kuala Lampur, Malaysia, 2001, vol. 1, pp. 355– 358. [11] R. Plamondon and G. Lorette, “Automatic signature verification and writer identification—The state of the art,” Pattern Recognit., vol. 22, no. 2, pp. 107–131, Jan. 1989. [12] Mingfu Zou, Jianjun Tong, Changping Liu and Zhengliang Lou,” Online Signature Verification Using Local Shape Analysis,” Proceedings of the Seventh International Conference on Document Analysis and Recognition (ICDAR 7), vol. 1, Edinburgh, U.K., Aug. 2003, pp. 314– 318. [13] Kholmatov A. and Yanikoglu B,” Biometric Authentication using Online Signatures,” Berlin, Springer, 3280, 2004. [14] Rioja F. R., Mariko N. M., Hector P. M. and Karina T. M, “Dynamic features Extraction for on-line signature verification,” Preceeding of the 14th international conference on electronic, communication and computers. IEEE, 156-161, 2004. Jyoti Bhalla, IJRIT

72

[15] Julian Fierrez-Aguilar, Loris Nanni, Jaime Lopez-Pe˜nalba, Javier Ortega-Garcia1 and Davide Maltoni,”An On-Line Signature Verification System Based on Fusion of Local and Global Information,” Springer Berlin / Heidelberg, vol: 3546/2005, pp. 523-532. [16] Marzuki Khalid, Hamam Mokayed, Rubiyah Yusof and Osamu Ono,”Online Signature Verification with Neural Networks Classifier and Fuzzy Inference,” Third Asia International Conference on Modeling & Simulation. IEEE, 236-241, 2009. [17] Christian Gruber, Thiemo Gruber, Sebastian Krinninger, and Bernhard Sick,” Online Signature Verification with Support Vector Machines Based on LCSS Kernel Functions,” IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 40, and NO. 4, pp. 1088-1100, 2010. [18] Ahmed Galib Reza, Hyotaek Lim, and Md Jahangir Alam,” An Efficient Online Signature Verification Scheme Using Dynamic Programming of String Matching, “Springer, LNCS, vol: 6935, pp. 590–597, 2011. [19] SVC2004 dataset that http://www.cse.ust.hk/svc2004/.

held

on

the

“First

International

Signature

Verification

Competition,”

[20] Mohammed J. Alhaddad, “Online Signature Verification using Probablistic Modelling and Neural Network” Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah, Saudi Arabia

Jyoti Bhalla, IJRIT

73