Hamid Amiri

Combination of local and global vision modelling for Arabic handwritten words recognition

We propose an Arabic handwritten word recognition system based on the idea of the PERCEPTRO system developed by Cote (Cote et al. (1998)) for Latin word recognition. It is a specific neural network, named transparent neural network, combining a global and a local vision modeling (GVM-LVM) of the word. In the forward propagation movement, the former (GVM) proposes a list of structural features characterizing the presence of some letters in the word. GVM proposes a list of possible letters and words containing these characteristics. Then, in the backpropagation movement, these letters are confirmed or not according to their proximity with corresponding printed letters. The correspondence between the letter shapes and the corresponding printed letters is performed by LVM using the correspondence of their Fourier descriptors, playing the role of a letter shape normalizer.

Discrimination between Arabic and Latin from bilingual documents

An important task in machine learning is the electronic reading of documents. In this process, discrimination between languages is one of the first steps in the problem of automatic document text recognition. We are interested in the processing of mixed Arabic/Latin printed documents. Our method is based essentially on the extraction of words. We first extract structural features of words and then recognize the writing language. We finally present the results of our classification approach and discuss possible improvements.

A new method for text detection and recognition in indoor scene for assisting blind people

Developing assisting system of handicapped persons become a challenging ask in research projects. Recently, a variety of tools are designed to help visually impaired or blind people object as a visual substitution system. The majority of these tools are based on the conversion of input information into auditory or tactile sensory information. Furthermore, object recognition and text retrieval are exploited in the visual substitution systems. Text detection and recognition provides the description of the surrounding environments, so that the blind person can readily recognize the scene. In this work, we aim to introduce a method for detecting and recognizing text in indoor scene. The process consists on the detection of the regions of interest that should contain the text using the connected component. Then, the text detection is provided by employing the images correlation. This component of an assistive blind person should be simple, so that the users are able to obtain the most informative feedback within the shortest time.

Integrate reasoning using OWL ontology and SWRL rules to online character recognition approach

The problem of recognition of handwritten characters is still an active area of research which accord more attention. In this context, our paper focuses on purpose of a new approach based on the integration of reasoning logic using OWL ontology and SWRL rules to the classification phase. Our ontology consists of a set of concepts and spatial relations between them, the main concept is the Arabic character which is segmented into fine grained concept named stroke.

A More Robust Mean Shift Tracker Using Joint Monogenic Signal Analysis and Color Histogram

This paper presents a robust object tracking method based on the methodologies of statistical texture analysis of 2D images based on the theory of monogenic signal analysis, jointed with the color histogram. This novel feature extraction method is embedded thereafter in the mean shift framework. Compared with methods of state-of-the-art mean shift trackers, this method proves to be more discriminant and less sensitive to noise. The experimental results proved that our proposed method can achieve robust tracking performances in complex situations with fewer mean shift iterations.

Local generic representation for patch uLBP-based face recognition with single training sample per subject

In this paper, we propose a novel paradigm of Patch uniform Local Binary Patterns (PuLBP) based Local Generic Representation (LGR) for face recognition. Indeed, we introduce a new block in which an uLBP is used to approximate both reference and variation subsets. Thus, we concentrate on the challenging problem of a single sample per person in a gallery set. Particularly, the main problem is whether only one training subject per class is available. One of the novelties of our technique is to generate virtual samples of each subject. The new sample generic image in a gallery set is adopted to produce the intra-personal variations of different individuals. We illustrate the experimental results of our new algorithm on different benchmark databases, including the AR face database, the Extended Yale B face database, the FRGC database and the FEI database.

Gabor-feature-based local generic representation for face recognition with single sample per person

This paper presents an approach called Gabor-feature-based Local Generic Representation (G-LGR), which take advantages of the sparse representation properties of face recognition in biometric applications. In this work, the main problem is that if only one training subject per class is available. One of the novelties of our new algorithm is to produce virtual samples of each subject; the new sample generic of a gallery set is used in order to generate the intra-personal variations of different individuals. We compare our approach against different state-of-the-art techniques using the AR face database.

Face Recognition with Single Training Sample per Subject.

This paper presents an approach called Patch uniform Local Binary Patterns (PuLBP) based Local Generic Representation (LGR) for face recognition. In fact, we insert a novel block that comports a uLBP in order to approximate both variation and reference subsets. Consequently, the focus will be on the difficult problem of a unique sample by person in a gallery set. More specifically, the major problem is if having solely one training person in every class is possible. The generation of virtual samples of every sample is one of the innovations of our technique. In a gallery set, each sample is used to generate the intra-personal variety of distinct individuals. We demonstrate the experimental results of our novel algorithm on many reference databases that include the FRGCv1, AR, the Georgia Tech (GT), the FEI and the Extended Cohn-Kanade.

The SOM tool in mechanical fault detection over an electric asynchronous drive

This paper presents a rotor bars Modeling. The rotor of an Electrical asynchronous machine is modeled by an equivalent electrical diagram related to the squirrel-cage connected together electrically and coupled magnetically, the frequencies characteristics of fault break bars. An intelligent strategy was adopted for fault detection in rotor using the map SOM (Self Organizing Map). It involves the most significant parameters of SOM, such as the topological structure of the map, the Kohonen learning algorithm, and also the activity diagram UML (Unified Modeling Language). Eventually, the measurement of the stator current on the experimental bench at a specific moment in the NDC (Non Destructive Control) Laboratory was applied. A comparative study of the fault detection performance was conducted under the SOM neural map and the spectral analysis method. It will be a more synthetic analysis.

A segmentation-free approach to strokes extraction from online isolated Arabic handwritten character

With the massive changes in input acquisition systems such as smart phones and tablets, the field of handwriting recognition has more attention accorded by a several researchers. This article addresses the problem of online Arabic character segmentation. Our approach is based on top-down segmentation-free of Arabic character by detecting the candidate points in the general chain code of the character to extraction strokes. The main goal of our work to ameliorate the recognition rate.