Writer-specific Dissimilarity Normalisation for Improved Writer-independent Off-line Signature Verification

In this paper we present a novel writer-independent off-line signature verification system. This system utilises the discrete Radon transform and a dynamic time warping algorithm for writer-independent signature representation in dissimilarity space. The system also considers writer-specific statistics for dissimilarity normalisation. A discriminant function, either linear or quadratic, is utilised for signature modelling and verification. We show that the feature extraction and dissimilarity representation framework proposed in this paper provides a successful platform for signature modelling and verification. We also show that the inclusion of writer-specific statistics during dissimilarity normalisation improves the proficiency of the proposed writer-independent verification system. When evaluated on Dolfing's data set, a signature database containing 1530 genuine signatures and 3000 amateur skilled forgeries, we show that the system presented in this paper outperforms existing systems also evaluated on this data set.