Bilal Hadjadji

The effective use of the one-class SVM classifier for handwritten signature verification based on writer-independent parameters

The limited number of writers and genuine signatures constitutes the main problem for designing a robust Handwritten Signature Verification System (HSVS). We propose, in this paper, the use of One-Class Support Vector Machine (OC-SVM) based on writer-independent parameters, which takes into consideration only genuine signatures and when forgery signatures are lack as counterexamples for designing the HSVS. The OC-SVM is effective when large samples are available for providing an accurate classification. However, available handwritten signature samples are often reduced and therefore the OC-SVM generates an inaccurate training and the classification is not well performed. In order to reduce the misclassification, we propose a modification of decision function used in the OC-SVM by adjusting carefully the optimal threshold through combining different distances used into the OC-SVM kernel. Experimental results conducted on CEDAR and GPDS handwritten signature datasets show the effective use of the proposed system comparatively to the state of the art. We propose handwritten signature verification for writer independent parameters.We propose to design HSVS system using only genuine signatures using OC-SVM.Applying a soft threshold in order to reduce the misclassification of the OC-SVM.Combination scheme is proposed through versus distances used into the OC-SVM kernel.Competitive results are obtained comparatively to the state of the art.

An Efficient Open System for Offline Handwritten Signature Identification based on Curvelet Transform and One-Class Principal Component Analysis

Abstract Offline handwritten signature identification has received less attention in comparison with the offline signature verification, despite its crucial applications such as in law-enforcements, automatic bank check and historical documents processing. In this paper, an Open Handwritten Signature Identification System (OHSIS) is proposed by using conjointly the Curvelet Transform (CT) and the One-Class classifier based on Principal Component Analysis (OC-PCA). CT is explored for feature generation due to its efficient characterization of curves contained into the local orientations within the signature image. While, OC-PCA is used for its effectiveness to absorb the high feature size generated by the CT and allows achieving at the same time an open system. Then, in order to improve the robustness of the OHSIS when few reference signatures are available, a new combination approach based on Choquet fuzzy integral is proposed to combine multiple individual OHSISs. Furthermore, a designing protocol with limited number of writers and reference signatures is proposed to perform a parameter-independent OHSIS. Experimental results conducted on standard CEDAR and GPDS handwritten signature datasets report 97.99% and 94.96% correct identification rate, respectively, which highlights the effectiveness of the proposed OHSIS since it can comfortably outperform the state-of-the-art when using few reference signatures.

Multiple One-Class Classifier Combination for Multi-class Classification

The One-Class Classifier (OCC) has been widely used for solving the one-class and multi-class classification problems. Its main advantage for multi-class is offering an open system and therefore allows easily extending new classes without retraining OCCs. However, extending the OCC to the multi-class classification achieves less accuracy comparatively to other multi-class classifiers. Hence, in order to improve the accuracy and keep the offered advantage we propose in this paper a Multiple Classifier System (MCS), which is composed of different types of OCC. Usually, the combination is performed using fixed or trained rules. Generally, the static weighted average is considered as straightforward combination rule. In this paper we propose a dynamic weighted average rule that calculates the appropriate weights for each test sample. Experimental results conducted on several real-world datasets proves the effective use of the proposed multiple classifier system where the dynamic weighted average rule achieves the best results for most datasets versus the mean, max, product and the static weighted average rules.

Fuzzy Integral Combination of One-Class Classifiers Designed for Multi-class Classification

One-Class Classifier (OCC) has been widely used for its ability to learn without counterexamples. Its main advantage for multi-class is offering an open system and therefore allows easily extending new classes without retraining OCCs. Generally, pattern recognition systems designed by a single source of information suffer from limitations such as the lack of uniqueness and non-universality. Thus, combining information from multiple sources becomes a mode for designing pattern recognition systems. Usually, fixed rules such as average, product, minimum and maximum are the standard used combiners for OCC ensembles. However, fixed combiners cannot be useful to treat some difficult cases. Hence, we propose in this paper a combination scheme of OCCs based on the use of fuzzy integral (FI) operators. Experimental results conducted on different types of OCC and two different handwritten datasets prove the superiority of FI against fixed combiners for an open multi-class classification based on OCC ensemble.

Two combination stages of clustered One-Class Classifiers for writer identification from text fragments

Abstract Writer identification based on handwritten fragments has been reported to give interesting performance. However, while the fragmentation process, inconsistent fragments are generated and affect badly the identification accuracy. Hence, in this paper, a clustered-based One-Class Classifier (OCC) is proposed in order to generate more robust classification model than the distance-based classifier for handwritten fragments. Besides, the problem of inconsistent fragments expands its effect to the test step. Thus, a Dynamic Fragment Weighting Combination (DFWC) rule is proposed to reduce the effect of inconsistent test fragments. Furthermore, due to the difficulty of performing a generic descriptor, three different descriptors related systems are designed and combined through an effective combination scheme based on Choquet fuzzy integral operator. Experimental results conducted on the well-known IFN/ENIT and IAM datasets show good adaptation of the OCC with DFWC. Moreover, the Choquet combination scheme offers more improvements to achieve 97.56% and 94.51% for the used datasets, respectively. The obtained results highlight the reliability of the proposed system in comparison with recent studies for writer identification issue.

Hybrid one-class classifier ensemble based on fuzzy integral for open-lexicon handwritten Arabic word recognition

One-class classifier (OCC) is involved for solving different kinds of problems due to its ability to represent a class distribution regardless the remaining classes. Its main advantage for multi-class classification is offering an open system and therefore allows easily extending new classes without retraining OCCs. So far, hidden Markov models, support vector machines and neural networks are the most used classifiers for Arabic word recognition, which provides a system with closed lexicon. In this paper, the OCCs are explored in order to perform an Arabic word recognition system with an open lexicon. Generally, pattern recognition systems designed by a single system suffer from limitations such as the lack of uniqueness and non-universality. Thus, combining multiple systems becomes an attractive research topic for performance and robustness enhancement. Fixed rules are commonly used us combiners for the hybrid OCC ensembles. The present paper aims to propose a combination scheme of OCCs based on the use of fuzzy integral (FI) operators. Furthermore, an alternative framework is proposed to design a parameter-independent and open-lexicon handwritten Arabic word recognition system as well as a new density measure function. Experimental results conducted on Arabic handwritten dataset using different types of OCCs with large number of classes highlight the superiority of FI for hybrid OCC ensembles.

Segmentation-verification based on fuzzy integral for connected handwritten digit recognition

This paper investigates a number of verification rules to validate the segmentation of connected handwritten digits. The verification technique based on statistical reasoning and fuzzy integrals is employed to verify the segmentation through decision functions produced by multiclass SVM based recognizers. The segmentation relies on an oriented sliding window which identifies potential cut points. The resulting segmented digits are fed to recognizers and the best segmentation is identified by the verification module that combines the recognizer outputs using fuzzy integrals. The proposed methodology is evaluated on a database of handwritten digits with single as well as multiple connections. Comparative analysis shows that the use of the fuzzy integral allows providing high recognition rates comparatively to the state of the art.

Keyword-guided Arabic word spotting in ancient document images using Curvelet descriptors

This paper deals with the contribution of Curvelet transform to generate more accurate word image descriptors for Arabic keyword spotting in ancient documents. Due to its properties, Curvelets can tolerate more scale distortions and more directional features in images. The process of Curvelet descriptor generation is applied to each word image in the dataset. Therefore, dynamic time warping algorithm is employed to match corresponding coefficients from Curvelet descriptor matrices. Experimental results on ancient Arabic document demonstrate that the characterization of the word image from the Curvelet descriptors offers better performance comparatively to the major state-of-the-art word image descriptors.