The effective use of the One-Class SVM classifier for reduced training samples and its application to handwritten signature verification

The One Class Support Vector Machine (OC-SVM) classifier has been used in many applications. Its main advantage is to train the classifier using only patterns belonging to the target class distribution. The OC-SVM is effective when large samples are available for providing an accurate classification. However, in some applications, as in handwritten signature verification, available handwritten signatures are often reduced and therefore the OC-SVM generates an inaccurate training and the classification is not well performed. In order to reduce the misclassification, we propose, in this paper, a modification of the decision function used in the OC-SVM by adjusting carefully the optimal threshold. Experimental results conducted on CEDAR and GDPS handwritten signature datasets show the effective use of the proposed method for reduced samples.