Mohamed Ali Mahjoub

Hidden Markov model for inferring user task using mouse movement

The assistive technology and e-learning have been widely used to improve web accessibility for disabled users. One of the issues of online web-based applications is to understand how a user interacts with online application and the strategy by which he reasons to perform a given activity. Know what the web user is doing can provide useful clues to better understand his behavior in order to guide him in his interaction process. This study proposes a methodology to analyze user mouse movement in order to infer the task performed by the user. To do this, a Hidden Markov Model is used for modeling the interaction of the learner with an e-learning application. The obtained results show the ability of our model to analyze the interaction in order to recognize the task performed by the learner.

Application of Bayesian networks for pattern recognition: Character recognition case

Pattern recognition is a wide field in progress. In particular, handwriting recognition has known a great development in the recent years. Several solutions have been directed towards the use of Bayesian networks, which have shown their ability to solve complex problems in many areas, and that is thanks to their ability to model inaccuracies, which are lacunae highly present in the manuscript field. In this paper, we recall the basics of these networks and the difficulties come across in their learning and inference algorithms to make a good decision. We present a state of using the BNs and especially RBDs in the pattern recognition and more exactly in the character recognition. We show, through the various considered works, the contribution of this technique in solving the limitations of the Markov models and its ability to represent efficiently the temporal notion and the dependencies between the variables during the writing process. Moreover, we retain the recorded limitations and some development perspectives.

Possibilistic reasoning effects on Hidden Markov Models effectiveness

Hidden Markov Models (HMM) have been widely used in classification tasks. Despite their efficiency in stochastic sequences labeling, they are overwhelmed by imperfect quality of used data in the learning and inference processes. In this paper, we try to evaluate the contribution of possibilistic theory in creating sequences of observations used by HMM models. Experimental results show that observation sequences, obtained by possibilistic reasoning significantly, improve the performance of HMM in the recognition of online e-learning activities.

Deep random forest-based learning transfer to SVM for brain tumor segmentation

Using neuroimaging techniques to diagnose brain tumors and detect both visible and invisible cancer cells infiltration boundaries motivated the emergence of diverse tumor segmentation algorithms. Noting the large variability in both tumor appearance and shape, the task of automatic segmentation becomes more difficult. In this paper, we propose a random-forest (RF) based learning transfer to SVM classifier method for segmenting tumor lesions while capturing their complex characteristics. Our framework is composed of two cascaded stages. In the first stage, we train a random forest to learn the mapping from the image space to the tumor label space. In the testing stage, we use the predicted label output from the random forest and feed it along with the testing intensity image to an SVM classifier to get the refined segmentation. Then we make our RF-SVM cascaded classification steps deep through an iterative process. We tested our method on 20 patients with high-grade gliomas from the Brain Tumor Image Segmentation Challenge (BRATS) dataset. Our proposed framework significantly outperformed SVM-based segmentation and RF-based segmentation-when used solely.

Characterization of Dynamic Bayesian Network The Dynamic Bayesian Network as temporal network

In this report, we will be interested at Dynamic Bayesian Network (DBNs) as a model that tries to incorporate temporal dimension with uncertainty. We start with basics of DBN where we especially focus in Inference and Learning concepts and algorithms. Then we will present different levels and methods of creating DBNs as well as approaches of incorporating temporal dimension in static Bayesian network.

Coupled Hidden Markov Model for video fall detection

Falls are a most common problem for old people. They can result in dangerous consequences even death. Many recent works have presented different approaches to detect fall and prevent dangerous outcomes. In this paper, we propose a coupled Hidden Markov Model (CHMM) for human fall detection from video streams. We use CHMM to model the motion and static spatial characteristic of human silhouette.

Tutorial and selected approaches on parameter learning in bayesian network with incomplete data

Bayesian networks (BN) are used in a big range of applications but they have one issue concerning parameter learning. In real application, training data are always incomplete or some nodes are hidden. To deal with this problem many learning parameter algorithms are suggested foreground EM, Gibbs sampling and RBE algorithms. This paper presents a tutorial of basic concepts and in particular techniques and algorithms associated with learning in Bayesian network with incomplete data. We present also selected applications in the fields.

Indexation of images of blood cells by generalized quaternary trees

In this paper, we are going to develop a system of indexation and research for the anomalies of red blood cells. This system uses a local approach which takes into account the spatial localization of globules in the images. The method is based on the image represented by generalized quaternary tree. In this approach we have given to the expert the possibility of choosing the adequate descriptor (color histogram, color texture), for every level, by a simple click on this last one. Currently, our system, called diagnosis, is capable of determining two anomalies of color type which can affected blood cells. He also allows to show the results via an effective graphic interface, which is simple to use and allows for a certain interaction with the user.

Indexing of ancient document images based on EM algorithm and tangent distance

In this paper we present a method of indexing of images from ancient documents based on Bayesian density estimation by the EM algorithm and tangent distance. Initially we present the procedure in case of known density of the mixture to discuss how to spend the density classification and therefore indexing. Once we cleared the problem justifies the choice of the density approximation by Gaussian mixture. Then we present the indexing algorithm based on the EM algorithm and the tangent distance. The tangent distance is a mathematical tool that compares two patterns (or images) by taking into account small transformations such as rotation and homothety, phenomena often encountered in ancient documents. The results show the robustness of the method compared to small global transformations.