Assessment of methods for image recreation from signature time-series data

Human signatures are widely used for biometric authentication. For automatic on-line signature verification, rather than storing an image of the completed signature, data is represented in the form of a time-series of pen position and status information allowing the extraction of temporal-based features. For visualization purposes, signature images need to be recreated from time-series data. In this paper we investigate the accuracy and verification performance of a series of interpolation methods for recreating a signature image from time-series data contained in two ISO/IEC data storage formats. Our experiments investigate dynamic data stored at various sample rates and signature images recreated at differing resolutions. Our study indicates possible best practice in terms of image recreation method, recreated image resolution, and temporal sample rate and assesses the effect of on the accuracy of reconstructed signature data.