Professional Documents
Culture Documents
ABSTRACT
1. INTRODUCTION
b) b)
Figure 4: a) Authentic signatures with quality mark Figure 5: Example of the output signal waveforms for au-
b) corresponding fake signatures thentic and fake signatures in Figure 4
The input and interface layer F1 consists of six sub- whole training procedure (the network places the template
layers (these are not illustrated in Figure 3); each sub-layer signatures only in one cluster and adapts the correspond-
has the same number of processing units as is the size of ing weights between F1 and F2 layers). When the train-
the feature vector. The purpose of these sub-layers is to ing is completed, the network is prepared for verification.
allow the ART-2 network to process continuously varying The parameters of F1 sub-layers are not changed during the
inputs. Moreover, they normalize the components of the verification, only the vigilance parameter ρ have to be set
feature vector and suppress the noise. The size of the F1 properly to the authentic and the fake signatures were set
layer (and hidden sub-layers) was 19 in our application. to right clusters. The three methods of setting the vigilance
The clustering layer F2 consists of two processing units parameter were tested in our work:
only, the former (labelled A) is active only if the feature vec-
tor corresponding to the authentic signature appears at the • manual setting M : vigilance parameter ρ is set to the
input of the network, the latter (labelled F) is active in other fixed value (ρ = 0.98) manually, for all verifications,
cases. More clusters are not enabled in our application.
• automatic setting A1 : ρA1 = min {ri } i = 1 · · · Nt ,
i
7. REFERENCES