Off-line Signature Verification Using Linear Regression Classifier
Bo Xu, Daozhi Lin, Hongyang Chao, Weifeng Li, Qingmin Liao · 2015
In this paper we propose a novel classification method based on Linear Regression Classification (LRC) for offline signature verification.The class-specific models can be simply established by using the registered samples (examples), and a test signature can be linearly represented by these registered samples.Then the tuned-LRC is constructed to capture the nonlinear information when the fundamental linear assumption is invalid in LRC.In contrast to the conventional classifiers used in signature verification, our proposed methods are very simple and no training stage is needed, and the dictionary can be easily expanded by additional samples.The experiments conducted on GPDS960Graysignature database demonstrate the effectiveness of the proposed methods.