Online Signature Verification Algorithm with a User-Specific Global-Parameter Fusion Model

Fusion is a promising strategy to improve performance in biometrics, and many fusion methods have been proposed. Most of them are user-generic fusion strategies, because generating user-specific strategies for each user is difficult. In this paper, we propose an online signature verification method using a user-specific global-parameter fusion model. The basic fusion model is a user-generic (global-parameter) fusion model, but by introducing a user-dependent mean vector, we can generate a user-specific fusion model. To evaluate the proposed algorithm, several experiments were performed by using three public databases. The proposed algorithm yielded equal error rates (EERs) of 4.0%, 8.6%, and 6.1% for the MCYT, SVC2004 task2, and MyIDea databases, respectively.