Mahmoud I. Khalil

A Database for Arabic Printed Character Recognition

Electronic Document Management (EDM) technology is being widely adopted as it makes for the efficient routing and retrieval of documents. Optical Character Recognition (OCR) is an important front end for such technology. Excellent OCR now exists for Latin based languages, but there are few systems that read Arabic, which limits the penetration of EDM into Arabic-speaking countries. In developing an OCR system for Arabic it is necessary to create a database of Arabic words. Such a database has many uses as well as in training and testing a recognition system. This paper provides a comprehensive study and analysis of Arabic words and explains how such a database was constructed. Unlike earlier studies, this paper describes a database developed using a large number of collected Arabic words (6 million). It also considers connected segments or Pieces of Arabic Words (PAWs) as well as Naked Pieces of Arabic Word (NPAWs); PAWS without diacritics. Background information concerning the Arabic language is also presented.

Building a multi-modal Arabic corpus (MMAC)

Traditionally, a corpus is a large structured set of text, electronically stored and processed. Corpora have become very important in the study of languages. They have opened new areas of linguistic research, which were unknown until recently. Corpora are also key to the development of optical character recognition (OCR) applications. Access to a corpus of both language and images is essential during OCR development, particularly while training and testing a recognition application. Excellent corpora have been developed for Latin-based languages, but few relate to the Arabic language. This limits the penetration of both corpus linguistics and OCR in Arabic-speaking countries. This paper describes the construction and provides a comprehensive study and analysis of a multi-modal Arabic corpus (MMAC) that is suitable for use in both OCR development and linguistics. MMAC currently contains six million Arabic words and, unlike previous corpora, also includes connected segments or pieces of Arabic words (PAWs) as well as naked pieces of Arabic words (NPAWs) and naked words (NWords); PAWs and Words without diacritical marks. Multi-modal data is generated from both text, gathered from a wide variety of sources, and images of existing documents. Text-based data is complemented by a set of artificially generated images showing each of the Words, NWords, PAWs and NPAWs involved. Applications are provided to generate a natural-looking degradation to the generated images. A ground truth annotation is offered for each such image, while natural images showing small paragraphs and full pages are augmented with representations of the text they depict. A statistical analysis and verification of the dataset has been carried out and is presented. MMAC was also tested using commercial OCR software and is publicly and freely available.

Expression and illumination invariant preprocessing technique for Face Recognition

Face recognition is one of the most intensively studied topics in the field of computer vision and pattern recognition. Facial expression, which changes face geometry, usually has an adverse effect on the performance of a face recognition system. Also, the appearance of a face image is usually affected by illumination conditions that hinder the automatic face recognition process. In this paper, we propose a robust expression and illumination invariant preprocessing technique for face recognition. We propose a new wavelet based method to eliminate the effect of expression variation and combine this method with an existing DCT based method for illumination normalization. Experimental results on the Yale database show that the proposed approach improves the face recognition rates.

Illumination invariant face recognition in logarithm Discrete Cosine Transform domain

Face is considered to be one of the biometrics in automatic person identification. For face recognition system to be practical, it should be robust to variations in illumination, pose and expression as humans recognize faces irrespective of all these variations. In this paper, we present an illumination invariant face recognition method in the logarithm Discrete Cosine Transform domain. We use an existing illumination normalization technique in the logarithm DCT domain. The main contribution in this paper is that we apply the Principal Component Analysis (PCA) algorithm for feature extraction in the DCT domain. By this, we skip the inverse DCT step and reduce the computational cost. Experimental results on the Yale B database show that we obtain the same results exactly as applying PCA in the spatial domain with the advantage of the reduced computational cost.

Neural Networks Pipeline for Offline Machine Printed Arabic OCR

In the context of Arabic optical characters recognition, Arabic poses more challenges because of its cursive nature. We purpose a system for recognizing a document containing Arabic text, using a pipeline of three neural networks. The first network model predicts the font size of an Arabic word, then the word is normalized to an 18pt font size that will be used to train the next two models. The second model is used to segment a word into characters. The problem of words segmentation in the Arabic language, as in many similar cursive languages, presents a challenge to the OCR systems. This paper presents a multichannel neural network to solve the offline segmentation of machine-printed Arabic documents. The segmented characters are then fed as an input to a convolutional neural network for Arabic characters recognition. The font size prediction model produced a test accuracy of 99.1%. The accuracy of the segmentation model using one font is 98.9%, while four-font model showed 95.5% accuracy. The whole pipeline showed an accuracy of 94.38% on Arabic Transparent font of size 18pt from APTI data set.

Epileptic seizure prediction using zero-crossings analysis of EEG wavelet detail coefficients

Predicting the occurrence of epileptic seizures can provide an enormous aid to epileptic patients. This paper introduces a novel patient-specific method for seizure prediction applied to scalp Electroencephalography (EEG) signals. The proposed method relies on the count of zero-crossings of wavelet detail coefficients of EEG signals as the major feature. This is followed by a binary classifier that discriminates between preictal and interictal states. The proposed method is practical for real-time applications given its computational efficiency as it uses an adaptive algorithm for channel selection to identify the optimum number of needed channels. Moreover, this method is robust against the variability across seizures for the same patient. Applied to data from 8 patients, the proposed method achieved high accuracy and sensitivity with an average accuracy of 94% and an average sensitivity of 96%. These results were obtained using only 10 minutes of training data as opposed to using hours of recordings typically used in traditional approaches.

Predictive Segmentation Using Multichannel Neural Networks in Arabic OCR System

This article offers an open vocabulary Arabic text recognition system using two neural networks, one for segmentation and another one for characters recognition. The problem of words segmentation in Arabic language, like many cursive languages, presents a challenge to the OCR systems. This paper presents a multichannel neural network to solve offline segmentation of machine-printed Arabic documents. The segmented characters are then used as input to a convolutional neural network for Arabic characters recognition. The accuracy of the segmentation model using one font is 98.9 %, while four-font model showed 95.5 % accuracy. The accuracy of characters recognition on Arabic Transparent font of size 18 pt from APTI data set is 94.8 %.

Fast Arabic Glyph Recognizer based on Haar Cascade Classifiers

Optical Character Recognition (OCR) is an important technology. The Arabic language lacks both the variety of OCR systems and the depth of research relative to Roman scripts. A machine learning, Haar-Cascade classifier (HCC) approach was introduced by Viola and Jones (Viola and Jones 2001) to achieve rapid object detection based on a boosted cascade Haar-like features. Here, that approach is modified for the first time to suit Arabic glyph recognition. The HCC approach eliminates problematic steps in the pre-processing and recognition phases and, most importantly, the character segmentation stage. A recognizer was produced for each of the 61 Arabic glyphs that exist after the removal of diacritical marks. These recognizers were trained and tested on some 2,000 images each. The system was tested with real text images and produces a recognition rate for Arabic glyphs of 87%. The proposed method is fast, with an average document recognition time of 14.7 seconds compared with 15.8 seconds for commercial software.

PHoG Features and Kullback-Leibler Divergence Based Ranking Method for Handwriting Recognition

Handwriting recognition is a research topic with a lot of challenges and complications. One of the main complications is big databases that affect classifier complexities and their ability to perform correctly. This paper introduces a new ranking approach that is proposed as a solution to this point of research. Per input test image, the approach sorts database classes from the nearest to the furthest based on the calculated ranks. Accordingly, the classification process is applied on only subset of best nearest neighbor classes rather than the whole database classes. The approach starts with computing simple regional-type features to group similar competitive database classes together using decision trees. This grouping process aims to split big database to multiple smaller ones. Decision trees match between test image and one of the constructed smaller databases. Finally, Kullback-Leibler divergence is measured between the pyramid histogram of gradients (PHoGs) features extracted from the test image and the members of the matched smaller database. This measurement sorts the matching classes to select smaller subset from them. This subset represents best nearest neighbors of test image that can be used for final classification. Reducing database size and focusing classification on subset of best nearest neighbor classes reduce the classifier complexity and increase the overall system classification accuracy. The proposed approach was applied on IFN-ENIT database, and its effect was tested on the SVM classifier.

Fast 3D tracking and quantization of small vascular structures in 3D medical images

This paper introduces an approach for fast tracking, quantization and centreline extraction of small vascular structures in medical imaging, with the problem of coronary segmentation used as an example. The tracking is achieved by propagating an explicit surface. This surface is represented as a triangular adaptive mesh where the element size is adjusted based on the current size of the vessel. Adaptive re-meshing is performed on-the-fly during propagation in an efficient manner. An effective self-intersection prevention method is introduced to address one of the major issues in triangular mesh offsetting.