Offline signature authentication: A back propagation-neural network approach

In the field o f Information Technology, security is the most important aspect to ensure confidentiality and avoid forgery. When we think about security, authentication is play an important role. To identify the authenticated person various biometric authentication techniques are used (like iris, fingers print, plum vain, signature authentication etc.). These techniques measures behavioral or physiological characteristics like a voice sample or a signature. In case of offline signature deals with the image of signature and the image is acquired by a digital camera or a scanner. In this case, the handwriting order, writing speed variation and skillfulness are the key points. Previously offline signature was verified u sing Hidden Markov Model (HMM), Support Vector Machine (SVM) or using some unique features of a signature. Some unique feature like global feature (like pixel density, pixel distribution and pixel axils), mask feature or grid feature. In any type of authentication technique three things are most important (i) False Acceptance Rate (FAR), (ii) False Rejection Rate (FRR) and (iii)Accuracy. In this paper, we deal with off-line signature and that signature was verified using Artificial Neural Network (ANN) u sing Back Propagation Neural Network; obtain satisfactory results when compared with existing approaches. Here after comparing target and predicted output, the error calculated is always less than 0.5.