Online Signature Verification Using Multi-Distance Measures and Weighting with Gradient Boosting

摘要

To improve performance of online signature verification while maintaining lower calculation cost and higher security, this study proposes a novel single-template strategy in function-based approaches. Specifically, we adopt dynamic time warping (DTW) barycenter averaging to obtain an effective mean template while preserving intra-user variability between all references. Then, by using the mean template, we calculate multi-distance measures: the multiple DTW from each feature with independent warping and the single DTW from all features with dependent warping. To boost the discriminative power, we apply a weighting scheme using gradient boosting to efficiently combine the multi-distance measures. The promising performance is demonstrated through its application to a popular SVC2004 Task2 dataset.