Synthetic generation of handwritten signatures based on spectral analysis

Abstract

A new method to generate synthetic online signatures is presented. The algorithm uses a parametrical model to generate the synthetic Discrete Fourier Transform (DFT) of the trajectory signals, which are then refined in the time domain and completed with a synthetic pressure function. Multiple samples of each signature are created so that synthetic databases may be produced. Quantitative and qualitative results are reported, showing that, in addition to presenting a very realistic appearance, the synthetically generated signatures have very similar characteristics to those that enable the recognition of real signatures.