A comparison between k-nearest neighbor and jk-nearest neighbor algorithms for signature verification

Signatures are widely used and accepted biometrics used for individual identification. Signatures are categorized as offline and online based on the input method. Online signatures contain more features than the regular offline signature, making them harder to forge. Several algorithms can be used for signature verification, such as the k-nearest neighbor. It is mainly used for one-class classification purposes. In this paper, both k-nearest neighbor and jk-nearest neighbor algorithms are presented, along with a comparison of both algorithms on online signature verification accuracy. The results are conducted using different combinations of verifiers using four different databases and showed that the jk-nearest neighbor classifier outperforms the traditional one-class k-NN classifier by 0.73-10% compared to a traditional one-class k-NN classifier.