k-NN based Writer Independent Offline Signature Verification System

Signature verification is a difficult research area since two people’s signatures may be similar, but an individual’s signature might vary depending on the situation. The accuracy of the signature verification framework is largely determined by the classifier and feature extraction scheme employed in the classification process. With this in mind, the current study looks into the effectiveness of the k-Nearest Neighbors classifier in conjunction with the Local Binary Pattern feature set for the development of a writer-independent offline signature verification system. To evaluate the system’s performance, two signature databases of 100 and 260 writers are used. Genuine signatures, as well as random forgery, unskilled forgery, and simulated forgery signatures, are considered for the development of the desired system, while genuine signatures, as well as random forgery, unskilled forgery, and simulated forgery signatures, are used to test the performance of the developed system. In simulation study false acceptance rate of 2.00%, 11.00% and 12.00% for random, unskilled, and simulated forgery signatures, respectively is obtained whereas the false rejection rate of 0.00% is achieved using Local Binary Pattern feature set.