A Single-Template Strategy Using Multi-Distance Measures and Weighting for Signature Verification
Manabu Okawa · 2018
In keeping with recent advances of artificial intelligence and relevant technical developments, online signature verification demands lower calculation cost and higher security, and, systems that meet these criteria without deteriorating performance are needed. In response to the described need, this study proposes a novel, single-template strategy for a function-based approach to online signature verification. Specifically, to obtain an effective single template while preserving intra-user variability between all reference samples, we adopt dynamic time warping (DTW) barycenter averaging. Then, by using the mean template, we calculate both DTW from univariate time series and DTW from multivariate time series. After that, to boost the discriminative power, we apply a weighting scheme using random forests to efficiently combine the two types of DTW. The experimental results confirm the effectiveness of the proposed method for online signature verification using the popular SVC2004 Task2 dataset.