4.2.1. First stage
Preclassi"caiton is the most important stage in signature
recognition. The success of recognition depends
heavily upon the success of "nding the group that matches
the input signature. From standard deviation results,
the slant feature appears powerful enough for use in a
"rst-stage classi"er. Experiments have shown that it is
quite e!ective to divide the data into four groups (negatively
(NS), vertically (VS), positively (PS) and horizontally
(HS) slanted) as a "rst stage resulting in 99%
accuracy. It has been found from the real-life data set that
it is very di$cult for the person to change the slant of his
or her hand while signing or even writing. In addition,
while negative slanting is predominant in Latin signatures,
it does not exist in Arabic signatures.