Identity Detection from On-line Handwriting Time Series

Identity detection is the key process behind any biometric authentication system. In this work a novel approach for identity detection from temporal information is proposed from the analysis of nonlinear time series. The handwriting can be considered as a manifestation of biomechanics. Observed timeseries from online handwritten signature is analysed to reconstruct the underlying dynamics behind handwriting. A similarity measure from the delay vector of reconstructed trajectories of handwriting time series, proposed earlier by the authors is studied here to evaluate its effectiveness in detecting identity of a person from his handwriting. Based on the proposed measure, an algorithm for discriminating genuine person from forger in authentication problem has been proposed. The simulation experiments have been done with SVC 2004 online handwriting signature data and the results show that the proposed approach is quite effective for biometric authentication.