Exploring feature variability in digitally captured signatures-Part I: Variability in genuine signatures and the effects of collection sessions and writing position.

Abstract

Digitally captured signatures (DCSs) pose new challenges for forensic handwriting examination compared with paper-based signatures. While DCSs provide spatial coordinates, time and force values for forensic examination, the variability of the features derived from these data remains insufficiently understood. This study therefore examined the within- and between-writer variability in 1560 genuine DCSs collected from 65 participants across eight sessions and three writing positions. These signatures were treated as questioned specimens and compared with 1820 reference DCSs collected later in one writing position. As expected, the results revealed substantial between-writer differences in absolute feature values, in variability, and in correlations despite general patterns across writers. Depending on the feature, between 55% (line length) and 96% (number of pen lifts) of the questioned signatures fell within their writer-specific reference ranges. Most values that were outside the range clustered around the reference distributions. This highlighted the limitations of interpretation based on a "within-outside the reference range" threshold. Across the eight sessions, only four out of the fifteen examined features (DTW scores for XY coordinates, time in air, total duration, and width) showed systematic changes, with questioned signatures becoming increasingly similar to the subsequently collected reference samples. Writing position also influenced several features significantly; for example, median force was lower when the pad was held in hand. The findings of this study underscore the importance of accounting for writer-specific feature variability and non-standard writing conditions when examining DCSs in forensic casework.