Investigating Long Short-Term Memory Networks for Various Pattern Recognition Problems

Otte, Sebastian; Liwicki, Marcus; Krechel, Dirk

doi:10.1007/978-3-319-08979-9_37

Cited by 10 publications

(6 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These results have outperformed the state-of-theart either for the case of using just one training signature (1vs1) [13] or the case of performing the average score of the four one-to-one comparisons (4vs1) [22]. In addition, it is important to highlight the results obtained in this work compared to the ones obtained by Otte et al in [10] where all experiments failed obtaining a 23.75% EER for the best case. In that work, standard LSTM architectures seemed not to be appropriate for the task of signature verification.…”

Section: Discussionmentioning

confidence: 46%

“…The LSTM RNNs proposed in that work seemed to authenticate genuine and impostor cases very well. However, as it was pointed out in [10], the method proposed in that work for training the LSTM RNNs is not feasible for real applications for various reasons. First, the authors considered the same users for both development and evaluation of the system.…”

Section: Introductionmentioning

confidence: 99%

“…Instead, they just reported the errors of the LSTM-outputs. In order to find some light on the feasibility of LSTM RNNs for signature verification purposes, Otte et al performed in [10] a deep analysis considering three different real scenarios: 1) training a general network to distinguish forgeries from genuine signatures on a large training set, 2) adopting a network that works perfectly on the training set to a specific writer, and 3) training the network on genuine signatures only. However, all experiments failed obtaining a 23.75% EER for the best configuration, far away from the best state-of-the-art results and concluding that LSTM RNN systems trained with standard mechanisms were not appropriate for the task of signature verification.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Biometric Signature Verification Using Recurrent Neural Networks

Tolosana

Vera-Rodríguez

Fiérrez

et al. 2017

2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

Architectures based on Recurrent Neural Networks (RNNs) have been successfully applied to many different tasks such as speech or handwriting recognition with state-of-theart 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.

show abstract

Section: Discussionmentioning

confidence: 46%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Biometric Signature Verification Using Recurrent Neural Networks

Tolosana

Vera-Rodríguez

Fiérrez

et al. 2017

2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)

View full text Add to dashboard Cite

show abstract

“…LSTM has shown to be successful in sequence prediction (Adi et al, 2016), sequence labeling (Sak and Beaufays, 2014), syntactic structure (Linzen et al, 2016) and long range semantic dependencies (He et al, 2017). LSTM's have an edge over CNN and Recurrent Neural Networks (RNN) in many ways (Otte et al, 2014). RNNs are effective when working with short term dependencies, but fail to recognize context or chronologically widely spaced input events or long-term dependencies.…”

Section: Long Short Term Memorymentioning

confidence: 99%

Machine Learning and Deep Learning for Phishing Email Classification using One-Hot Encoding

Nandi¹,

Bagui²,

White³

2021

Journal of Computer Science

View full text Add to dashboard Cite

Representation of text is a significant task in Natural Language Processing (NLP) and in recent years Deep Learning (DL) and Machine Learning (ML) have been widely used in various NLP tasks like topic classification, sentiment analysis and language translation. Until very recently, little work has been devoted to semantic analysis in phishing detection or phishing email detection. The novelty of this study is in using deep semantic analysis to capture inherent characteristics of the text body. One-hot encoding was used with DL and ML techniques to classify emails as phishing or nonphishing. A comparison of various parameters and hyperparameters was performed for DL. The results of various ML models, Naïve Bayes, SVM, Decision Tree, as well as DL models, Convolutional Neural Networks (CNN) and Long Short Term Memory (LSTM), were presented. The DL models performed better than the ML models in terms of accuracy, but the ML models performed better than the DL models in terms of computation time. CNN with Word Embedding performed the best in terms of accuracy (96.34%), demonstrating the effectiveness of semantic analysis in phishing email detection.

show abstract

“…One of the first studies that analysed the potential of current deep learning approaches for on-line signature verification was [21]. In that work, Otte et al performed an exhaustive analysis of Long Short-Term Memory (LSTM) RNNs using a total of 20 users and 12 genuine signatures per user for training.…”

Section: On-line Signature Verificationmentioning

confidence: 99%

DeepSign: Deep On-Line Signature Verification

Tolosana,

Vera-Rodriguez,

Fierrez

et al. 2020

Preprint

View full text Add to dashboard Cite

Deep learning has become a breathtaking technology in the last years, overcoming traditional handcrafted approaches and even humans for many different tasks. However, in some tasks, such as the verification of handwritten signatures, the amount of publicly available data is scarce, what makes difficult to test the real limits of deep learning. In addition to the lack of public data, it is not easy to evaluate the improvements of novel proposed approaches as different databases and experimental protocols are usually considered.The main contributions of this study are: i) we provide an indepth analysis of state-of-the-art deep learning approaches for on-line signature verification, ii) we present and describe the new DeepSignDB on-line handwritten signature biometric public database 1 , iii) we propose a standard experimental protocol and benchmark to be used for the research community in order to perform a fair comparison of novel approaches with the state of the art, and iv) we adapt and evaluate our recent deep learning approach named Time-Aligned Recurrent Neural Networks (TA-RNNs) 2 for the task of on-line handwritten signature verification. This approach combines the potential of Dynamic Time Warping and Recurrent Neural Networks to train more robust systems against forgeries. Our proposed TA-RNN system outperforms the state of the art, achieving results even below 2.0% EER when considering skilled forgery impostors and just one training signature per user.

show abstract

Investigating Long Short-Term Memory Networks for Various Pattern Recognition Problems

Cited by 10 publications

References 17 publications

Biometric Signature Verification Using Recurrent Neural Networks

Biometric Signature Verification Using Recurrent Neural Networks

Machine Learning and Deep Learning for Phishing Email Classification using One-Hot Encoding

DeepSign: Deep On-Line Signature Verification

Contact Info

Product

Resources

About