| Summary: | 碩士 === 元智工學院 === 電機與資訊工程研究所 === 84 === Signing is a ballistic motion that involves an unconscious
reflex action. The writer is generally thinking about what heis
signing rather than how to spell his name or form thatwriting.
Because of the difference in musculature, individualvariations
in dynamic characteristics of handwriting lead oneto expect that
it is possible to detect forgeries. An accurateand efficient
similarity measure algorithm is proposed in thisstudy that can
be used to check whether a test signature isgenuine or not. By
applying the transformation processdeveloped in [1], the spatial
difference of signatures thatcaused by natural variability is
corrected roughly, and thecorresponding stroke segments between
the test signature and thereference signature are matched
properly. In order to obtainmore accurate results, the curve
matching algorithm proposedby Tseng [2] is modified to calculate
the similarity measurefor each matched stroke segment pair. Both
of the static anddynamic similarities of the two signatures are
calculatedutilizing the Fourier descriptors. These values will
be used tocheck whether the test signature is genuine or not.
Theadvantages of the method includes: (1) comparing both of
thestatic and the dynamic signals of the signatures can increase
the stability of the system; (2) calculating the similarities
for each matched stroke segment pair will increase the
accuracyof the system; and (3) applying the same similarity
computationalgorithm to both the static and the dynamic signals
makes thesystem more efficient.
|
|---|
