Although the computerized verification of handwritten signatures has been extensively
studied in the past three decades, the results on the field are still unsatisfactory.
This paper provides a generic scientific approach for modeling different kinds of
signature verification systems and estimating their performance limitations based
on simple properties of the signature database, which is used to evaluate the system.
Although our current approach concentrates only on skilled forgeries, it could be
well extended to handle unskilled and random forgeries and thereby provide a model,
which is applicable to current commercial signature verification systems.