Biometric Signature Verification Using Recurrent Neural Networks

TL;DR

This paper explores the use of LSTM-based Siamese RNNs for on-line signature verification, demonstrating significant improvements over previous methods on the BiosecurID dataset.

Contribution

It introduces a Siamese LSTM RNN architecture for signature verification and shows its effectiveness on a real-world dataset, outperforming existing approaches.

Findings

Achieved 17.76% to 28.00% relative verification improvement.

Validated the feasibility of RNNs for practical signature verification.

Demonstrated superior performance on the BiosecurID benchmark.

Abstract

Architectures based on Recurrent Neural Networks (RNNs) have been successfully applied to many different tasks such as speech or handwriting recognition with state-of-the-art results. The main contribution of this work is to analyse the feasibility of RNNs for on-line signature verification in real practical scenarios. We have considered a system based on Long Short-Term Memory (LSTM) with a Siamese architecture whose goal is to learn a similarity metric from pairs of signatures. For the experimental work, the BiosecurID database comprised of 400 users and 4 separated acquisition sessions are considered. Our proposed LSTM RNN system has outperformed the results of recent published works on the BiosecurID benchmark in figures ranging from 17.76% to 28.00% relative verification performance improvement for skilled forgeries.