An optimal approach to aligning on-line handwritten signatures

On-line handwritten signature aligning is crucial in verification. This paper deals with this aligning as an optimization problem. By transforming the squared Euclidean distance based DTW distance in terms of geometric transformations including rotation, scaling and translation, corresponding optimal objective functions are formulated and solved in closed form. Subsequently, an iterative optimization procedure, named DTWOA, is proposed as a systematic aligning approach. Experimental results conducted on SVC2004 dataset show that DTWOA converges rapidly, and the standard deviation of the optimal DTW distances is much smaller and smoother than the original, which manifests the availability of our work.