Signature verification based on machine learning and deep learning techniques: A review

An essential forensic way to determine if such a particular autograph seems to be genuine or counterfeit termed as sign authentication. For instance, organizations are facing several kinds of signature forgery as tracing of another person’s signature, duplication, imitation documents and random forgery which leads to major issues. In many economic, administrative, universities perhaps other professional contexts, it’s indeed critical to avoid documentation forgery. Hence, the authors analyzed various techniques as how machine learning and deep learning techniques appropriate in detecting signature counterfeit. Moreover, this review focused on analyzing kinds of signature images, data, handwritten or digital, offline or online signature and framework done by existing investigators. In addition, the authors made comparisons on machine learning techniques to measure performance for verifying signature in terms of accuracy, FNR, FPR and computational time. Similarly, deep based neural network approaches are compared for evaluating performance based on training and testing accuracy and other neural network parameters namely batch size, learning rate and epochs. Finally, the authors evaluated overall analysis in signature verification both online, offline as well.