Maher Khemakhem

A Database for Arabic Handwritten Text Image Recognition and Writer Identification

Standard databases play essential roles for evaluating and comparing results obtained by different groups of researchers. In this paper, an Arabic Handwritten Text Images Database written by Multiple Writers (AHTID/MW) is introduced. This database can be used for research in the recognition of Arabic handwritten text with open vocabulary, word segmentation and writer identification. The AHTID/MW contains 3710 text lines and 22896 words written by 53 native writers of Arabic. In addition, ground truth annotation is provided for each text image. The database is freely available for worldwide researchers.

A novel approach for off-line Arabic writer identification based on stroke feature combination

This paper presents a novel approach for off-line text-independent Arabic writer identification. The approach operates in four steps: 1) handwritten text is segmented into strokes after an image thinning step; 2)length, height/width ratio and curvature stroke features are extracted; 3) five feature vectors are computed: stroke length/ratio probability distribution function (PDF), stroke length/ratio horizontal and vertical cross-correlation, stroke length/curvature PDF, stroke length/curvature horizontal and vertical cross-correlation, and stroke length/curvature and length/ratio cross-correlation; 4) classification is carried out using different metrics and the Borda count ranking algorithm. A first experimental evaluation performed on 40 writers from the IFN/ENIT database produced a promising identification rate of 92.5% for Top1 and 100% for Top5.

Database for Arabic Printed Text Recognition Research

This paper presents a real database for the Arabic printed text recognition, APTID / MF (Arabic Printed Text Image Database / Multi-Font).This database can be used to evaluate the system that recognizes Arabic printed texts with an open vocabulary. APTID / MF may be also used for research in word segmentation and font identification. APTID / MF is obtained from 387 pages of Arabic printed documents scanned with grayscale format and 300 dpi resolutions. From this documents, 1,845 text-blocks have been extracted. In addition ground truth file is provided for each texts-block. APTID / MF also includes an Arabic printed character image dataset made up of 27,402 samples. The database is freely available to interested researchers.

A multipurpose multi-agent system based on a loosely coupled architecture to speedup the DTW algorithm for Arabic printed cursive OCR

Summary form only given. This paper aims to prove through an analytical study that even a loosely coupled architecture such as a grid computing can be used to support an effective multiagent system in many ways especially to speedup the dynamic time warp algorithm (DTW) for Arabic printed cursive OCR. The adaptation of the (DTW) algorithm to Arabic printed cursive OCR provides excellent recognition rate without a prior characters segmentation especially for some font sizes (16, 18, and 20). Experiments show that the DTW algorithm is able to recognize perfectly Arabic printed cursive characters by the use of a references library formed only of isolated characters (one occurrence per character). The major problem of the DTW is the slowness of its response time because of the enormous amount of computation to achieve. Multiagent systems (MAS) are very flexible and are considered to be a very attractive framework to support distributed applications, performance evaluation of the proposed MAS shows that the task granularity is one of the key performance parameters that can govern the number of agents to use and the achievable speedup. We found in particular that if the amount of computation achieved by the MAS is greater 2 times than the maximum amount of communication and synchronization required by any two agents (of the MAS) then the proposed MAS is effective.

An Effective Combination of MPP Contour-Based Features for Off-Line Text-Independent Arabic Writer Identification

This paper proposes an off-line, text-independent, Arabic writer identification approach, using a combination of probability distribution function (PDF) features. In writer identification, the success of PDFs in terms of homogeneity, classification and identification rates encouraged researchers to study them with different types of structural features. Intensive experiments achieved on 82 writers from the IFN/ENIT database show, in particular, that 6 simple feature vectors based on the length, direction, angle and curvature measurements, which are extracted from the minimum-perimeter polygon (MPP) contours of the pieces of Arabic words, can be used to reach promising Arabic writer identification rates. The results are obtained using a set of distance metrics and the Borda ranking algorithm for classification, and the best identification rates are 90.2% for top1, and 97.5% for top10, which confirm the consistency of the proposed approach.

Arabic Writer Identification and Verification Using Template Matching Analysis of Texture

This paper introduces an original approach for off-line, text-independent Arabic writer identification and verification. Indeed, the handwriting texture is inspected in this approach relatively to its multi-resolution aspects of directionality, angularity and curvature. The 10 resulting feature vectors are classified using common distance metrics, and then combined using a simple and experimentally proven weighting scheme. The combining of different types of textural feature vectors in our approach allowed the computer to capture efficiently the writers individual and unique handwriting characteristics. Intensive experiments were conducted with 557 handwriting samples from 100 different writers in the IFN/ENIT database, and we obtained promising writer identification rates of 85% for Top1, 90% for Top2 and 95% for Top10. As for writer verification, an acceptable equal error rate (EER) of 5.9% ± 0.11% was obtained.

Towards an agent based framework for the design of secure web services

Web services (WS) are commonly provided by software components accessed through remote procedure calls. One of the major problems encountered during the access process to any WS is their security in terms of, for instance, execution and exchanged data. Several solutions and proposals have been made to solve this problem mainly by OASIS. Despite their contributions, current solutions and proposals do not completely solve the security problem inherent to WS. In fact, they focused only on the definition of the security rules at the implementation level and neglected the design level of WS, which can induce important added values. In this paper, we outline how the agent based design approach can be exploited to provide for safe WS both at the design and implementation levels. More specifically, we show how volatile, mobile agents can guarantee secure WS through their autonomy, intelligence and mobility. In addition, we illustrate how various security solutions can be integrated in this design approach.

Mitigation of Sybil Attacks in Structured P2P Overlay Networks

Sybils, i.e., the bane of malicious identities under a common control entity, are commonly controlled by an attacker. In Sybil attack, a single malicious user forges multiple fake identities and pretends to be multiple, distinct physical node in the system. However, Sybil attack is one of the most dangerous attacks that plague current structured Peer-to-Peer overlay networks. This attack is employed to target honest peers and hence subvert the system. In this paper, we describe a new methodology to analyze, detect, and mitigate Sybil attacks. We examine in detail this attack, the most challenging problem that currently propagates in structured Peer-to-Peer overlay networks. We identify features and search for behavioral attributes that may serve to identify such attacks. We were able to analyze them in depth, using honey pots, which allows us to collect information to distinguish Sybil nodes from honest nodes. Furthermore, we introduce a mitigation strategy that alleviates some of the effects of such an attack by injecting some commands into the Sybil node or subvert the communication channel.

Agent technology based resources management in Cloud Computing

This paper proposes a model based on agent technology for the optimization of resources management in Cloud Computing. We focus indeed on the providers side in order to provide them a tool which is able to optimize automatically their resources management in a customized manner since we observed the lack of such a tool and this to the best of our knowledge.

Energy management strategy in cloud computing: a perspective study

Optimizing energy consumption in cloud computing is yet a challenge despite the diversity of the proposed energy management strategies. Indeed, and during our related work study we have observed that the different elements or components which should be considered in order to be able to properly manage energy consumption in a cloud computing context are not well defined and/or discussed in terms of importance. This makes the proper classification and/or comparison of the different proposed strategies or techniques very difficult. Consequently, this paper aims, on the one hand, at defining and discussing properly such components in order to create a guideline and, on the other hand, to ease both the classification and the comparison of these proposed strategies and techniques. Second and after discussing some common weaknesses related to the current energy consumption optimization techniques and methods, this paper proposes energy-saving technique which uses a novel load detecting policy. This policy is based on the median absolute deviation method which uses the median and the standard deviation to calculate upper and lower thresholds which aim to classify hosts into either overloaded or under-loaded state. Simulation results have shown better results of the proposed technique compared to the existing ones especially in reducing energy consumption and the number of virtual machine migrations in addition to better active host time. Indeed, we found that the average of saved energy is around 40% compared to the built in techniques in cloudSim.