Online signature verification using hybrid transform features
Andile Miaba, Mandlenkosi Victor Gwetu, Serestina Viriri · 2018
This study is based on online signature verification using hybrid transform features that are gathered from dynamic signature signals. The signatures are transformed to digital signals in order to model them in alternative domains for the emphasis of salient features. Transforms used in this study are: Discrete Fourier, Discrete Cosine and Discrete Wavelet Transforms. Although these transforms are normally used in digital signal processing for tasks such as speech and audio filtering, they are also applicable to other domains such as image processing and financial modelling. The goal of this study is to investigate the effect of combining transform features to authenticate signatures. Due to genuine human error and lack of consistency, comparing signatures requires preprocessing to assist with standardization. The proposed approach achieved a FAR and FRR of 0.655 and 0.235, respectively on the SVC2004 database. The results confirm the proposed approach as competitive and promising.