Handwriting features based detection of fake signatures
Anton Akusok, Leonardo Espinosa-Leal, Kaj-Mikael Björk, Amaury Lendasse, Renjie Hu · 2021
Detection of fake signatures is a hard task. In this paper, we present a novel method for detecting trained forgeries using features extracted from sliding windows with different overlaps on a public available dataset of static images of signatures. Using a linear machine learning model named Extreme Learning Machine (ELM), our methodology achieves, in average, an Equal Error Rates (EER) of 2.31% for an overlap of 90%. In line with the state-of-the-art results available in the scientific literature.