Image-based handwritten signature verification using hybrid methods of discrete Radon transform, principal component analysis and probabilistic neural network

Ooi, Shih Yin; Teoh, Andrew Beng Jin; Pang, Ying Han; Hiew, Bee Yan

doi:10.1016/j.asoc.2015.11.039

Cited by 79 publications

(17 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multi-script learning approaches could be considered to improve performance on all scripts. Lastly, tables 5, 6, 7 and 8 compare the results we obtained with SPP-Fixed (considering EER user thresholds [40] 10 Contours (chi squared distance) 6.44 Wen et al [41] 5 RPF (HMM) 15.02 Vargas et al [19] 5 LBP (SVM) 11.9 Vargas et al [19] 10 LBP (SVM) 7.08 Ooi et al [42] 5 DRT + PCA (PNN) 13.86 Ooi et al [42] 10 DRT + PCA (PNN) 9.87 Soleimani et al [39] 5 HOG (DMML) 13.44 Soleimani et al [39] 10 HOG (DMML) 9.86 Hafemann et al [6] 10 1 morphology (SVM) 11.81 Kumar et al [45] 1 Surroundness (NN) 8.33 Bharathi and Shekar [46] 12 Chain code (SVM) 7.84 Guerbai et al [12] 4 Curvelet transform (OC-SVM) 8.7 Guerbai et al [12] 8 Curvelet transform (OC-SVM) 7.83 Guerbai et al [12] 12 Curvelet transform (OC-SVM) 5.6 Hafemann et al [6] 12 SigNet-F (SVM) 4.63 (± 0.42) Present Work 10 SigNet-SPP-300dpi 3.60 (± 1.26) Present Work 10 SigNet-SPP-300dpi (finetuned) 2.33 (± 0.88) using genuine signatures and skilled forgeries) with the state-of-the-art in GPDS, MCYT, Cedar and Brazilian PUC-PR, respectively. We observe results competitive to the state of the art in all datasets.…”

Section: Resultsmentioning

confidence: 99%

Fixed-sized representation learning from offline handwritten signatures of different sizes

Hafemann

Oliveira

Sabourin

2018

IJDAR

View full text Add to dashboard Cite

This is a pre-print of an article published in the International Journal on Document Analysis and Recognition. The final authenticated version is available online at: https://doi.org/10.1007/s10032-018-0301-6.Abstract Methods for learning feature representations for Offline Handwritten Signature Verification have been successfully proposed in recent literature, using Deep Convolutional Neural Networks to learn representations from signature pixels. Such methods reported large performance improvements compared to handcrafted feature extractors. However, they also introduced an important constraint: the inputs to the neural networks must have a fixed size, while signatures vary significantly in size between different users. In this paper we propose addressing this issue by learning a fixed-sized representation from variable-sized signatures by modifying the network architecture, using Spatial Pyramid Pooling. We also investigate the impact of the resolution of the images used for training, and the impact of adapting (fine-tuning) the representations to new operating conditions (different acquisition protocols, such as writing instruments and scan resolution). On the GPDS dataset, we achieve results comparable with the stateof-the-art, while removing the constraint of having a maximum size for the signatures to be processed. We also show that using higher resolutions (300 or 600dpi) can improve performance when skilled forgeries from a subset of users are available for feature learning, but lower resolutions (around 100dpi) can be used if only L. Hafemann · R. Sabourin Laboratoire d'imagerie, de vision et d'intelligence artificielle (LIVIA) Ecole de technologie supérieure, 1100, rue Notre-Dame Ouest, genuine signatures are used. Lastly, we show that finetuning can improve performance when the operating conditions change.

show abstract

Section: Resultsmentioning

confidence: 99%

Fixed-sized representation learning from offline handwritten signatures of different sizes

Hafemann

Oliveira

Sabourin

2018

IJDAR

View full text Add to dashboard Cite

show abstract

“…Ooia (2016) [4] used Radon Transform for feature extraction from images. Discrete Radon Transform (DRT) signifies the image projection at various angles.…”

Section: Feature Extractionmentioning

confidence: 99%

Design of Offline and Online Writer Inference Technique

2019

IJITEE

View full text Add to dashboard Cite

Writer inference systems tend to identify and verify the authorship of the handwritten documents. Each writer will have his own style of writing that uniquely identifies the writer. Hence authorship identification finds its application in forensic document analysis. It is also considered as one of the biometric features of a person, so helps in security to uniquely identify a person. Recognition of writers online has its application in detecting the identity thefts. That is compromising one’s social media account and sending messages to others as if he were an authentic sender. By discriminating the writing characteristics of the original and intruder, the masquerader can be identified. In this survey various works contributing to feature extraction and prediction of writers are discussed.

show abstract

“…Among them, maximum amount of charge at one time and total amount of charge can be found in the design, while horizontal distance and elevation difference should be measured in site. However, many researchers have pointed out that this is not enough to accurately forecast the PPV [26,27]. Instead of using regression coefficient, more measured data should be adopted.…”

Section: Dataset and Dimensionalitymentioning

confidence: 99%

“…In recent years, a huge number of researchers have performed a train of works on combination of ANNs and principal component analysis [4,6,16,[19][20][21][22][23][24][25][26][27][28][29][30][31]. Nevertheless, the application of Mean Impact Value and Factor Analysis is exactly rare in previous research.…”

Section: Introductionmentioning

confidence: 99%

Prediction of Peak Velocity of Blasting Vibration Based on Artificial Neural Network Optimized by Dimensionality Reduction of FA-MIV

Zhang

Jin

2018

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Blasting vibration is harmful to the nearby habitants and dwellings in diverse geotechnical engineering. In this paper, a novel scheme based on Artificial Neural Network (ANN) method optimized by dimensionality reduction of Factor Analysis and Mean Impact Value (FA-MIV) is proposed to predict peak particle velocity (PPV) of blasting vibration. To construct the model, nine parameters of field measurement are taken as undetermined input parameters for research, while peak particle velocity (PPV) is considered as output parameter. With the application of FA, common factors are extracted from undetermined input parameters. Then, principal components are defined as a linear combination of common factors. The weight of each principal components effected on output parameter is ranked according to the calculation of MIV, and two principal components with minimum weight are eliminated. Ultimately, output parameter (PPV) is explained in a low-dimensional space with four input characteristic parameters. In the prepared database consisting of 108 datasets, 98 datasets are used for the training of the model, while the rest are used for testing performance. The performances of the ANN models are compared with regression analysis, in terms of coefficient of determination (R2) and mean absolute error (MAE). It is found that the performances of ANN models with using FA-MIV are superior to those of models without using FA-MIV in the prediction of PPV. In addition, the abilities of ANN models are all superior to regression analysis in the prediction of PPV. The result obtained from ELM is more accurate than BPNN and MVRA models.

show abstract

Image-based handwritten signature verification using hybrid methods of discrete Radon transform, principal component analysis and probabilistic neural network

Cited by 79 publications

References 30 publications

Fixed-sized representation learning from offline handwritten signatures of different sizes

Fixed-sized representation learning from offline handwritten signatures of different sizes

Design of Offline and Online Writer Inference Technique

Prediction of Peak Velocity of Blasting Vibration Based on Artificial Neural Network Optimized by Dimensionality Reduction of FA-MIV

Contact Info

Product

Resources

About