Hassanin M. Al-Barhamtoshy

A large vocabulary system for Arabic online handwriting recognition

The success of using Hidden Markov Models (HMMs) for speech recognition application has motivated the adoption of these models for handwriting recognition especially the online handwriting that has large similarity with the speech signal as a sequential process. Some languages such as Arabic, Farsi and Urdo include large number of delayed strokes that are written above or below most letters and usually written delayed in time. These delayed strokes represent a modeling challenge for the conventional left-right HMM that is commonly used for Automatic Speech Recognition (ASR) systems. In this paper, we introduce a new approach for handling delayed strokes in Arabic online handwriting recognition using HMMs. We also show that several modeling approaches such as context based tri-grapheme models, speaker adaptive training and discriminative training that are currently used in most state-of-the-art ASR systems can provide similar performance improvement for Hand Writing Recognition (HWR) systems. Finally, we show that using a multi-pass decoder that use the computationally less expensive models in the early passes can provide an Arabic large vocabulary HWR system with practical decoding time. We evaluated the performance of our proposed Arabic HWR system using two databases of small and large lexicons. For the small lexicon data set, our system achieved competing results compared to the best reported state-of-the-art Arabic HWR systems. For the large lexicon, our system achieved promising results (accuracy and time) for a vocabulary size of 64k words with the possibility of adapting the models for specific writers to get even better results.

Arabic document layout analysis

Document layout analysis is a key step in the process of converting document images into text. Arabic language script is cursive and written in different styles which cause some challenges in the analysis of Arabic text documents. In this paper, we introduce an approach for Arabic documents layout analysis. In that approach, the document is segmented into set of zones using morphological operations. The segmented zones are classified as text or non-text ones using a support vector machine classifier. Features used in zone classification are combination between texture-based features and connected component-based features. The textural-based feature vector size is reduced using genetic algorithm. Classified text zones are clustered, using adaptive sample set clustering algorithm, into lines. Each segmented line is segmented into words by clustering inter- and intra-spaces. The proposed system was evaluated against two other systems that represent the best available tools for the Arabic documents analysis, and evaluation results show that the proposed system works well on multi-font and multi-size documents with a variety of layouts even on some historical documents.

A Holistic Technique for an Arabic OCR System

Analytical based approaches in Optical Character Recognition (OCR) systems can endure a significant amount of segmentation errors, especially when dealing with cursive languages such as the Arabic language with frequent overlapping between characters. Holistic based approaches that consider whole words as single units were introduced as an effective approach to avoid such segmentation errors. Still the main challenge for these approaches is their computation complexity, especially when dealing with large vocabulary applications. In this paper, we introduce a computationally efficient, holistic Arabic OCR system. A lexicon reduction approach based on clustering similar shaped words is used to reduce recognition time. Using global word level Discrete Cosine Transform (DCT) based features in combination with local block based features, our proposed approach managed to generalize for new font sizes that were not included in the training data. Evaluation results for the approach using different test sets from modern and historical Arabic books are promising compared with state of art Arabic OCR systems.

A Toolkit for Teaching Arabic Handwriting

percentage of people who produce a neat and clear handwriting is declining sharply. The traditional approach for handwriting teaching is to have a dedicated teacher for long hours of handwriting practice. Unfortunately, this is not feasible in many cases. In this paper we introduce an automated tool for teaching Arabic handwriting using tablet PCs and on-line handwriting recognition techniques. This tool can simulate the tasks performed by a human handwriting teacher of detecting the segments of hypothesized writing errors and producing instructive real time feedback to help the student to improve his handwriting quality. The tool consists of two main components, the guided writing component and the free writing component. In the guided writing mode the student is required to write over transparent images for the training examples to limit his hand movements. After the student acquires the basic skills of handwriting he can practice the free writing mode where he writes with his own style, as he usually does in his daily handwritings. The first version of the tool was tested in several schools for children with edge ranging 4-11. The results are promising and show that this tool can help students to analyze their own writing and understand how they can improve it.

Arabic Spoken Language Identification System (ASLIS): A Proposed System to Identifying Modern Standard Arabic (MSA) and Egyptian Dialect

There are millions of people in the world speak many languages. To communicate with each other it is necessary to know the language which we use. To do this operation we use language identification system.

Toward cloud-based mixed reality e-learning system

This paper presents design and implementation of a mixed reality e-learning architecture integrated with cloud computing technology. Consequently, the paper discusses a cloud based mixed reality computing framework for e-learning collaboration model. The proposed framework for the cloud-based mixed reality is designed to open the mixed reality dataset courses and related contents. The framework employees’ four modules, each module implements various services related to this module. Also, the framework provides an environment of online mixed reality learning courses. The structure of the framework includes (1) learner, (2) mixed reality, (3) learning objects and styles, and (4) manager with user friendly modules. Consequently, ways of deploying such cloud computing can be defined as public, private or hybrid followed by the security and privacy techniques to protect the proposed framework. So, confidentiality, integrity and availability of material of courses will satisfied, to examine the threats and security issues for the cloud computing framework.

Diagnosis of Dyslexia using computation analysis

This paper introduces to diagnosis of Dyslexia using computing system, considered people difficulties in reading, spelling, writing and speaking. Consequently, a computational analysis classifier will be achieved using dyslexia metrics techniques. Accordingly, Gibson test of brain skills will be used with effect of working memory, auditing (hearing and speech) and visual memory and cognition, visual and auditing perceptions, writing and motor skills, math and time management, behavior, health, development and personality, cognitive ability in peoples with learning specially reading difficulties taken into our consideration. Computation analysis with classifiers will be used to analyze the proposed dataset that includes 80 children records. This computation model is designed and implemented to help uncover the underlying problems that may affect learning to read or write as well as problems that may also cause issues with memorizes comprehension. This model is implemented to help counselors and parents understand the difficulty and get kid in the correct passageway to education success.

Towards large scale image similarity discovery model

The paper introduces to the digitization and features extraction processes of large volume of imaging documents and stored as images using mechanisms of big data and cloud technology. So, layout analysis, image representation, feature extraction and transformation huge amount of the prepared document images are presented in this paper. Accordingly, an efficient way reliable and highly clustering functionality of these document images will be focused. Consequently, image-based extractor using “document image similarity” is the main methodology to apply this paper. Many tasks have been proposed to contribute such idea and also to support retrieval performance. Different methods such as layout analysis, image representation, features vectors creation, similarity measure, accuracy computation and document image retrieval will be presented.

A fedearted E-learning cloud system based on mixed reality

The continuous evolution of new technologies and devices encourage researchers and specialists, in learning and E-learning domain, to take advantage of them in order to enhance learning and E-learning tasks. Due to its flexibility and some other benefits compared to the conventional learning, E-learning gained a lot from the field of learning and training. Cloud computing presents today a very important low-cost infrastructure which can respond to all kinds of software and hardware user needs. In this paper, we propose a federated E-learning cloud system capable of enhancing, substantially, the quality of E-learning by using the mixed reality technology. Actually, our system is intended to be used by any learner including those having some special needs (talent and smart peoples). One of the attractive sides of the proposed system is its capability to use existing courses and learning objects provided on the Internet in order to deliver them to the end-user in a customized manner. The proposed system is intended to be built on a federated cloud infrastructure as a set of ubiquitous services that can be accessed by any user.

Universal metadata repository for document analysis and recognition

Document Analysis and Recognition (DAR) has two main objectives, first the analysis of the physical structure of the input image of the document, which should lead to the correct identification of the corresponding different homogeneous components and their boundaries in terms of XY coordinates. Second, each of these homogeneous components should be recognized in such a way that, if it is a text image, consequently this image should be recognized and translated into an intelligible text. DAR remains one of the most challenging topics in pattern recognition. Indeed, despite the diversity of the proposed approaches, techniques and methods, results remain very weak and away from expectations especially for several categories of documents such as complex, low quality, handwritten and historical documents. The complex structure and/or morphology of such documents are behind the weakness of results of these proposed approaches, techniques and methods. One of the challenging problems related to this topic is the creation of standard datasets that can be used by all stakeholders of this topic such as system developers, expert evaluators, and users. In addition, another challenging problem is how one could take advantages of all existing datasets that unfortunately are dispersed around the world without knowing, most of the times, any information about their locations and the way to reach them. As an attempt to solve the two mentioned above problems, we propose in this paper a Universal Datasets Repository for Document Analysis and Recognition (UMDAR) that has, in fact, a twofold advantage. First, it can help dataset creators to standardize their datasets and making them accessible to the research community once published on the proposed repository. Second, it can be used as a central which bridges in a smart manner between datasets and all DAR stakeholders.