On The Effects of Sampling Rate and Interpolation in HMM-Based Dynamic Signature Verification

In this work, resampling techniques and interpolation are applied to on-line signatures. Their effect on the performance of an on-line signature system using hidden Markov models is studied. The presented techniques are based on linear interpolation and Catmull-Rom cubic splines. Experimental results are provided on the MCYT database comprising 16,500 signatures from 330 subjects and as many skilled forgeries. Our approach allows to reduce the sampling rate of the on-line signature capture system, leading to the reduction of storage resource requirements and increased simplicity without compromising and even enhancing the system performance. This also leads to the discussion of which is the minimum sampling rate for HMM-based dynamic signature verification.