Mouldi Bedda

Handwritten Arabic character recognition based on SVM Classifier

This paper describes new methods for handwritten Arabic character recognition. We propose a novel algorithm for smoothing image and segmentation of the Arabic character using width writing estimated from skeleton character. The moments and Fourier descriptor of profile projection and centroid distance are used as features of each character these feature are invariant in translation , rotation and scale we apply Principal component Analysis (PCA) as data processing algorithm to features vector in order to reduce dimension. The classifier proposed in this work is based on Support Vector Machines (SVM) wich considerd an recent optimal classifier up to now. The results show that these methods are very powerful for isolated handwritten Arabic character.

Probabilistic classification based on copula for speech recognitation: an overview

Copula functions have been widely used in economics and finance and more recently it has been used in same field of pattern recognition. This paper presents an overview for introducing the use of Copulas function to supervised probabilistic classification applied to automatic speech recognition.

Using Active Learning in Text Classification of Quranic Sciences

The key idea behind active learning is that if the learning method is allowed to choose the data to learn from, the amount of data needed for the training phase can be significantly reduced. Thus, the cost of manual annotating the data will be less, and the process of learning can be accelerated. Most of the studies on applying active learning methods to automatic text classification focused on requesting the label of a single unlabeled document in each iteration. Unlike English, There are very few researches done in this area for the Arabic text. In this paper, we present a novel active learning method for Arabic text classification using multi-class SVM. The proposed method selects a batch of informative samples for manually labeling by an expert. The experimental results show that employing our method can significantly reduce the need for labeled training data.

Arabic text categorization using SVM active learning technique: An overview

Support vector machine is one of the famous techniques used in active learning to reduce the data labeling effort in different fields of pattern recognition. Most of the studies on applying active learning methods to automatic text classification focused on requesting the label of a single unlabeled document in each iteration. In this paper, we present a novel batch mode active learning using SVM for Arabic text classification.

Spoken Arabic Digits Recognition Based on (GMM) for E-Quran Voice Browsing: Application for Blind Category

People with low or no visual ability must also be able to manipulate, operate and browse the electronic reading devices of the Quran by a simple use of the voice (operation known as Voice-In/Voice-Out). The main operations of navigation and exploration of these devices, as the movement between verses or between pages can be fully realized through a voice recognition system of Arabic numbers. In this paper, we propose the use of voice recognition of Arabic digits as a way to use these devices, for this purpose, we present the method of speech recognition based on: (GMM) classifier, known for its effectiveness and scalability in speech modeling and the leading approach in speech recognition feature extraction Delta-Delta Mel-frequency cepstral coefficients (MFCC). The experimental results with the obtained parameters demonstrate the effectiveness of the digit recognition on a dataset in 99.31% of cases, which is highly satisfactory compared to previous works on spoken Arabic digits speech recognition.

Realization of an offline system for the recognition of the handwritten numeric chains

In this study we propose an off line system for the recognition of the handwritten numeric chains. Firstly, we realized a recognition system of the isolated handwritten digits, in this part, the study is based mainly on the evaluation of neural network performances, trained with the gradient back propagation algorithm and fed by several parameter vectors; the objective of this operation is to determine the parameter vector who gives the best performances. The used parameters to form the parameter vectors are extracted on the binary images of the digits by several methods: the distribution sequence, sondes application, the Barr features, The centered moments of image coding according to the Freeman directions, and the centered moments of the different projections and profiles. Secondly, the study is extended for the reading of the handwritten numeric chains constituted of a variable number of digits. Vertical projection was used to segment the numeric chain at isolated digits and every digit (or segment) was presented separately to the entry of the system achieved in the first part (recognition system of the isolated handwritten digits).

Recognition of the Arabic Handwritten Words of the Algerian Departments

In the context of the handwriting recognition, we propose an off line system for the recognition of the Arabic handwritten words of the Algerian departments. The study is based mainly on the evaluation of neural network performances, trained with the gradient back propagation algorithm. The used parameters to form the input vector of the neural network are extracted on the binary images of the handwritten word by several methods. The Distribution parameters, the centered moments of the different projections of the different segments, the centered moments of the word image coding according to the directions of Freeman, and the Barr features applied binary image of the word and on its different segments. The classification is achieved by a multi layers perceptron. A detailed experiment is carried and satisfactory recognition results are reported.