Intra-class variation representation for on-line signature verification using wavelet and fractal analysis

Signature is an important legal personal identification. Selecting a good feature representation is a significant step in designing a signature verification system. Single resolution function approach used in on-line signature verification faces the difficulty in identifying the intra-class variations of the features extracted. Such an approach might cause the acceptance of forged signatures that have similar patterns as the original and the rejection of genuine signatures that have high intra-class variations. This paper discusses the intra-class variation representation in on-line signature verification using wavelet and fractal analysis. With the achievement performance of an average improvement of 18% in genuine test verification rate and 7% in forged test verification rate compared to the single resolution function approach, it proves that the intra-class variations are important for on-line signature verification.