Regresar

State of the art: Humans performance contributions in signature recognition via crowdsourcing and manual annotation

Abstract:

This work visualizes a state - of - the - art study of researches in the biometry specially in signature recognition, to show the potential of collaborative tools such as crowdsourcing and a tool for human - assisted schemes to improve automatic signature recognition systems. We present an analysis of experiments of evaluation of signatures made through crowdsourcing and labeling of attributes inspired by the FDE analysis. These experiments allow us to establish human performance in signature recognition tasks, the same ones that they use, an HTML interface through the Mturk platform, a public BiosecureID database, and an interface developed in Matlab for manual annotation. The results of these studies demonstrate how human intervention can help improve the performance of automatic signature recognition systems and be able to propose a semi-automatic system schema.