Off-Line Signature Verification using Higher-Order Local Autocorrelation Features

摘要

A critical issue in the field of off-line signature verification is the choice of a good feature set Signatures of the same person are relatively stable over time and are reasonably invariant in rotation and in scale, when they are written down on a pre-established frame like a bank check. However invariance in translation generally does not occur. Therefore an automatic signature verification system will either provide a translation correction mechanism, or will alternatively use translation invariant features. This work follows the second alternative and proposes the use of Higher-Order Local Autocorrelation (HLAC) features for off-line signature verification. A set of experiments were carried to investigate the potential of HLAC features as global signature descriptors. A database with 40 signatures examples of 20 signers were built for the evaluation. The measured false rejection and false acceptance rates are consistent with the reported performance other systems which also use global descriptors.