Keywords image processing, signature verification, false acceptance rate, false rejection rate, forgeries, feature vectors, support vector machines.

Transcription

1 International Journal of Computer Application and Engineering Technology Volume 3-Issue2, Apr 2014.Pp OFFLINE SIGNATURE VERIFICATION SYSTEM -A REVIEW Pooja Department of Computer Engineering, Yadavindra College of Engineering, Talwandi Sabo, Punjab, India E.mail: Rajan Goyal Assistant Professor, Department of Computer Engineering, Yadavindra College of Engineering, Talwandi Sabo, Punjab, India E.mail: Abstract The person s signature is an important biometric attribute which is used for authentication and identification of individual. Today, Signature verification is one of the most important features. There are many parameters for security checking like password, pin code, but signature recognition is the very popular because it is quite accurate and cost efficient too. On the other hand, it s very difficult to remember pin code or password. It helps to reduce frauds in Banks, Business and improves customer service and security. In this paper, we present the review of Offline Signature Verification System and different Feature Vectors and Classifiers. Keywords image processing, signature verification, false acceptance rate, false rejection rate, forgeries, feature vectors, support vector machines. 1. INTRODUCTION Signature verification is a very important for person authentication. It is widely used in the fields of finance, access control and security. Signature verification is the process which is carried out to determine whether a given signature is genuine or not. Signature verification is different from character recognition thus consider signature as a complete image with some particular curves that represent a particular writing style of the person. There are two basic types of signature verification. One is Online Signature Verification and other is Offline Signature Verification. On-line approach uses an electronic tablet and a stylus connected to a computer to extract information about a signature and takes dynamic information like speed of writing for verification purpose.in Off-line signature verification, it is hard to segment signature strokes due to highly stylish writing styles and signatures are captured by camera or scanner. In off-line signature verification system features are extracted from scanned signature image. 2. SIGNATURE VERIFICATION CONCEPTS 2.1. Types of Forgies Random: These are written by those person who doesn t know the shape of the of the original signature. Simple: These signatures are are written by those person who knows how the signature looks like by knowing the name of the signer without much practice. Skilled: written by Person who knows exactly how the original Signature looks like with proper practice. 188

2 2.2. Types of Features Features extracted for Static signature verification can be broadly divided into three main types: Global features- These are extracted from the whole signature image. Global features include signature area, signature height to width ratio, Centre of gravity etc. Local features- These are extracted from the small portion of signature image. These features are computationally expensive but are much more accurate than global features. Local features include local pixel density, slant features, critical points etc. Geometric features It represent the characteristic geometry of a signature that keeps both their global as well as local feature properties Error Rate False Rejection Rate (FRR): This is the percentage of genuine signatures rejected as false. FRR = 100 False Acceptance Rate (FAR): This is the percentage of false signatures accepted as genuine. FAR = PROCESSING STEPS IN STATIC SIGNATURE VERIFICATION Figure 1. General overview of Static signature verification system Image Acquisition Preprocessing Feature Extraction Verification Classification Image acquisition The signatures to be processed by the system should be in the digital image format. The data for the offline signature verification system acquire from camera or scanner. Preprocessing The purpose of Signature pre-processing step is to make signatures standard and ready for feature extraction. Preprocessing is a necessary step to improve the accuracy of Feature extraction and Verification. Feature Extraction The efficiency of a signature verification system mainly depends on Feature extraction stage. Feature extraction techniques should be fast and easy to compute so that system has low computational power. Verification This step compares test signature features with genuine signature features based on various pattern classification techniques and makes a final decision for verification as genuine or forged signature. 189

3 4. METHODS FOR STATIC SIGNATURE VERIFICATION Euclidean Distance Model- Euclidean Distance Model is used for classification. This classifier is good for features extracted and fast in computation. This is the simple distance between a pair of vectors of size n. We can calculate distance using Euclidean distance model by following equation. In threshold calculation these distances are useful. Let X (x1, x2..xn) and Y(y1,y2.yn) are two vectors of size n. We can calculate distance (d) by using equation. distance d x y Banshider Majhi, Y Santhosh Reddy, D Prasanna Babu [4] proposed an Off-line Signature Verification based on novel features. Authors used Euclidean Distance Model for classification. Features are extracted based on vertical splitting and horizontal splitting. Authors reported that 9.75% acceptance rate for simple forgeries and 16.36% acceptance rate for skilled forgeries and 14.58% rejection rate for genuine signatures. Mrs. Tulsi Gupta [10] developed Off-line Signature Verification. Four features centroid, trisurface, six-fold surface and shape number feature are extracted using Euclidean Distance Model. Authors reported FAR of 43.6% and FRR of 38.1%. Neural Networks An NN is a parallel computing system that consists of a large number of simple processors with many interconnections. The main reasons for usage of neural networks are their power and ease of use. Neural Networks are highly suited to modeling global aspects of handwritten signatures. H. Baltzakisa, N. Papamarkos [5] developed a signature verification technique based on a two-stage neural network classifier in which global, grid and texture features are extracted. For each one of these feature sets a special two stage Perceptron OCON (one-class-one-network) classification structure was implemented. In the first stage, the classifier combines the decision results of the neural networks and the Euclidean distance obtained using the three feature sets. The performance of the system was checked by the use of the remaining subset (TS) of 500 signatures. A FAR of 9.81%, FRR of 3% and an overall efficiency of 90.09% was achieved. Support Vector Machines V.Vapnik et al. introduced this new learning method. SVMs are machine learning algorithms for binary classification based on recent advances in statistical learning theory. S. Audet, P. Bansal, and S. Baskaran [12], designed Off-Line Signature Verification using Virtual Support Vector Machines. They used global, directional and grid features of signatures. Virtual Support Vector Machine (VSVM) was used to verify and classify the signatures and FAR of 16.0% and FRR of 13.0% was obtained. Ahmed Abdelrahman, Ahmed Abdallah [1] developed Signature Verification System Based on Support Vector Machine Classifier. In this paper, Global features are extracted from the signatures using radon transform. Authors used a database of 2250 signatures (genuine signatures and skilled forgeries) from 75 writers in the proposed signature verification system a performance of approximately 82% is achieved. Template Matching Techniques Template matching is one of the earliest and simplest approaches to pattern recognition. A pattern class is represented by a template. Such a template pattern can either be a curve or an image. Dynamic Time Wrapping is the most popular template matching technique for Static signature verification. A. Piyush Shanker and A. N. Rajagopalan 190

4 Table 1. Performance evaluation of different methods Serial No. Method FAR (%) FRR (%) 1. Support Vector Machine [8] Two Stage Neural Network Classifier [5] Enhanced Modified Direction Feature [11] Based on Fuzzy Modeling [15] Virtual Support Vector Machine [12] Modified Dynamic Time Wrapping [2] Hierarchical Random Graph Model [9] Euclidean Distance Model [10] Weighting Factor Based Approach [6] Hybrid Statistical Modelling [14] [2] proposed a signature verification system based on Modified Dynamic Time Warping (DTW). Authors reported that with a threshold value of 1.5, the system has close to 0% acceptance rate for casual forgeries, 20% acceptance rate for skilled forgeries, and about 25% rejection rate for genuine signatures. 5. PERFORMANCE EVALUATION WITH RESULTS The performance of system is determined based on the accuracy of classification between the genuine and forged signature. Evaluation parameters for any signature verification system are FAR and FRR. The performances of different methods with results are shown in Table CONCLUSION This review article presents a brief overview of the recent works on Static signature verification. Different existing approaches used for signature verification are discussed and compared along with their FAR and FRR. The results shows that the accuracy of existing available signature verification systems is not enough to implement in public use thus more research on Static Signature verification is required. There are still many challenges in this domain which includes the signatures from the same person are similar but not identical. Person s signature often changes because of age, illness, geographic location and up to some extent the emotional state of the person. Thus there is a need to combine different classifiers with different feature vectors in future work to enhance performance. 7. REFERENCES [1] Ahmed Abdelrahman, Ahmed Abdallah, Signature Verification System Based on Support Vector Machine Classifier, The International Conference On Information Technology, [2] A. Piyush Shanker and A. N. Rajagopalan, Off-line signature verification using DTW, Pattern Recognition Letters, Vol. 28, pp: , [3] Ashwini Pansare, Shalini Bhatia, Handwritten Signature Verification using Neural Network, International Journal of Applied Information Systems, Vol. 1, pp: 44-49, [4] Banshider Majhi, Y Santhosh Reddy, D Prasanna Babu, Novel Features for Off-line Signature Verification, International Journal of Computers, Communications & Control, Vol. 1, pp: 17-24,

5 [5] H. Baltzakis, N. Papamarkos, A new signature verification technique based on a two-stage neural network classifier, Elsevier, pp: , [6] Ji Jun-wen, Chen Chuan-bo, Chen Xiao-su Off-line Chinese Signature Verification: Using Weighting Factor on Similarity Computation, (ICEBISS-2010), pp: 1-4, [7] Kishor T. Mane, Vandana G. Pujari, Signature Matching with Automated Cheque System, International Conference on Intelligent Systems and Signal Processing, pp: , [8] Mandeep Kaur Randhawa,A.K.Sharma,R.K Sharma, off-line signature verification based on hu s moment invariants and zone features using support vector machine, International Journal Of Latest Trends In Engineering And Technology, Vol. 1(3), pp: 16-23, [9] M. Piekarczyk, Hierarchical Random Graph Model for Off-line Handwritten Signatures Recognition, ICCISIS, pp: , [10] Mrs. Tulsi Gupta, Off-line Signature Verification, International Journal of Computer Application, Vol. 3(2), pp: , [11] S. Armand, M. Blumenstein, V. Muthukkumarasamy, Off-line Signature Verification using the Enhanced Modified Direction Feature and Neural-based Classification, (IJCNN-2006), pp: , [12] S. Audet, P. Bansal, and S. Baskaran, Off-line signature verification using virtual support vector machines, ECSE 526 Artificial Intelligence,pp: 1-8, [13] Sayantan Roy, Sushila Maheshkar, Offline Signature Verification using Grid based and Centroid based Approach, International Journal of Computer Applications, Vol. 86, pp: 35-39, [14] S.M.S.Ahmad, A.Shakil, M.A.Faudzi, R. M. Anwar, M.A.M.Balbed A Hybrid Statistical Modelling, Normalization and Inferencing Techniques of an Off-line Signature Verification System, pp: 6-11, [15] T. Wei Q. Yizheng, M. Zhiqiang, Off-line Signature Verification Based on Fuzzy Modelling with the Optimal Number of Rules, IEEE International Conference on Multimedia and Expo, pp: , [16] Vaishali M. Deshmukh, Sachin A. Murab, Signature Recognition &Verification Using ANN, International Journal of Innovative Technology and Exploring Engineering, Vol. 1(6), pp: 6-8,