Mohamed Saber Naceur

New biometric approach based on geometrical humain brain patterns recognition: Some preliminary results

In this paper, we describe a new biometric approach based on geometrical characteristics of brain shape. Specifically, we use these geometrics characteristics as a biometric feature to identify individuals. For this purpose, Magnetic Resonance Imaging (MRI) images are considered. We show that using a single slice from an MRI volumetric image, acquired at a given level, one can extract many significant geometrical descriptors related to inter-individual variability of brain shape that can be also used to identify individuals. Explicitly, the proposed biometric approach combines two main phases. In the first phase, features extraction (FE) are achieved in order to obtain brain geometrical descriptors vector, called in this work GDB vector. A second phase is called similarity measurement (SM). Finally, the proposed algorithm is evaluated on the Open Access Series of Imaging Studies (OASIS) database containing brain MRI Images. Results using 220 classes show that high accuracy of 98.76% to identify individuals are obtained.

3D Tumor Segmentation from Volumetric Brain MR Images Using Level-Sets Method

1.1 Statement of the problem Segmentation in volumetric images is a tool allowing a diagnostics automation and as well will assist experts in quantitative and qualitative analysis. It’s an important step in various applications such as visualization, morphometrics and image-guided surgery. In the context of neuro-imaging, brain tumor segmentation from Magnetic Resonance Images (MRI) is extremely important for treatment planning, therapy monitoring, examining efficacy of radiation and drug treatments and studying the difference between healthy subjects and subjects with brain tumor. The task of manually segmentation of brain tumor from MR images is generally time-consuming and difficult. Anyway, the task is done by marking by hand the tumor regions slice-by-slice which generates set of jaggy images, so the practitioner is confronted with a succession of boundary which he mentally stacked up to be made a 3D shape of brain tumor. This shape is inevitably subjective and becomes infeasible when dealing with large data sets, also there is losing of information in the third dimension because is not taken into account in the segmentation process. All this, affect the quality and accuracy of clinical diagnosis. An automatic or semi-automatic segmentation method of brain tumor that takes entire information within the volumetric MR image into account is desirable as it reduces the load on the human raters and generates optimal segmented images (Wang & al., 2004), (Michael & al., 2001), (Lynn & al., 2001). Specially, automatic brain tumor segmentation presents many challenges and involves various disciplines such us pathology, MRI physics and image processing. Brain tumors are difficult to segment because they vary greatly in size and position, may be of any size, may have a variety of shapes and may have overlapping intensities with normal tissue and edema. This leads to numerous segmentation approaches of automatic brain tumor extraction. Low-level segmentation methods, such as pixel-based clustering, region growing, and filter-based edge detection, requires additional pre-processing and post-processing as well as considerable amounts of expert intervention and a priori knowledge on the regions of interest (ROI) (Sahoo & al., 1988). Recently, several attempts have been made to apply deformable models to brain image analysis (Moon & al., 2002). Indeed, deformable models refer to a large class of computer vision methods and have proved to be a successful segmentation technique for a wide range of applications. Deformable models, on the other hand, provide an explicit representation of the boundary and the ROI shape. They combine several desirable features such as inherent connectivity and smoothness, which counteract

Data analytics for predictive maintenance of industrial robots

The predictive maintenance of industrial machines is one of the challenging applications in the new era of Industry 4.0. Thanks to the predictive capabilities offered by the emerging smart data analytics, data-driven approaches for condition monitoring are becoming widely used for early detection of anomalies on production machines. The aim of this paper is to provide insights on the predictive maintenance of industrial robots and the possibility of building a condition-monitoring system based on the data analysis of robots power measurements. A predictive modeling approach is proposed to detect robot manipulator accuracy errors based on robots current data analysis for predictive maintenance purposes. An experimental procedure is also carried out to oversee the correlation between the robot accuracy error and a set of extracted features from current time-series, and to evaluate the proposed predictive modeling. The obtained results are satisfactory and prove the feasibility of building a data-driven condition monitoring of robot manipulators using the electrical power time-series data analysis.

Automatic detection of slow-wave sleep and REM-sleep stages using polysomnographic ECG signals

We describe in this paper a new approach of classifying the different sleep stages only by focusing on the polysomnographic ECG signals. We show the pre-processing technique of the ECG signals. At the same time the identification and elimination of the different types of artifacts which contain the signal and its reconstruction are shown. The automatic classification of the slow-deep sleep and the rapid eye movement sleep called in this work REM-sleep consists in extracting physiological indicators that characterize these two sleep stages through the polysomnographic ECG signal. In other words, this classification is based on the analysis of the cardiac rhythm during a nights sleep.

Using brain prints as new biometric feature for human recognition

Abstract Considering the evolution of neuroimaging in the medical field, some new emerging biometric modalities become interesting and promising candidates to recognize persons. These modalities are considered as a part of “Hidden Biometrics” which consists in using clinical measurements and medical imaging for recognition purpose. The main motivation in using hidden biometrics is the fact that system attacks may be extremely difficult to consider. This specificity highly contributes in increasing the robustness in terms of person verification and identification. In this paper, we deal with a novel non-invasive approach to recognize persons by extracting a brain signature, called “brainprint”. In particular, we explored the brain cortical regions of volumetric brain MRI (Magnetic Resonance Imaging) images, acquired from 220 healthy subjects. For each subject, four 3D cortical surfaces are considered, then transformed into 2D cortical folds maps. From the resulting textures, brainprints are constructed by extracting features using Wavelet Gabor Transform. These brainprints are considered in this work as a discriminative signature of the brain. In terms of performance evaluation, we show that an EER = 2.90 ± 0.47 is reached for verification mode. On the other hand, when dealing with identification, the proposed approach allows a recognition rate of 99.6%.

A comparison of classification methods for analysis of remotely sensed hyperpectral data

This paper examines the utility of hyperspectral imagery for remote sensing data analysis. The acquired data volumes are very important, often reaching hundreds or thousands of channels for a single scene observed. Certainly, the large quantity contained in the hyperspectral database is accompanied by a complex physical content and consequently a considerable time computing which can affect the quality of treatment. These channels come from a very fine spectral sampling, allow discriminating and differentiating constituents that are spectrally close. Furthermore, hyperspectral imaging is a technique that is as strong potential initiating new research in the development of mathematical morphology. This theory, mainly inspired by the image processing problems, extends to a new more complex scope of which is hyperspectral mathematical morphology. Applied to hyperspectral data in hyperdimensional features spaces, we compare two proposed classification approaches. The first method is based on centralized segmentation methodology which exploits the information complementarities. The second method is based on hierarchical clustering which consists on combining a divisive clustering approach applied at a high-level and an agglomerative clustering operating at a low-level.

Selection of hyperspectral bands by adopting a dimension reduction strategy for recognition of multispectral palmprint

Palmprint is a unique and reliable biometric characteristic with high usability. Many works have been carried out on this field, during the past decades. Different algorithms and systems have been proposed and built successfully. Multispectral or hyperspectral palmprint imaging and recognition can be a potential solution to these systems because it can acquire more discriminative information for personal identity recognition. The Selection of the spectral bands is the most important step to develop the multispectral palmprint system. Most of the work done is based on methods by choosing the selected bands empirically. This work represents a preliminary study on the selection of bands by analyzing hyperspectral palmprint data (900nm ∼ 1600nm). We use an hyperspectral data provided by the “GPDShandsSWIRhyperspectral”. We adopted a dimension reduction strategy for the recognition of multispectral palmprint. We conducted a comparative study between two methods using the algorithms SOBI and JADE for the reduction of size bands. The results obtained showed that we can reduce the 20 bands chosen to 16 bands without having to modify the information from the image.

High- and low-level hierarchical classification algorithm based on source separation process

Abstract. High-dimensional data applications have earned great attention in recent years. We focus on remote sensing data analysis on high-dimensional space like hyperspectral data. From a methodological viewpoint, remote sensing data analysis is not a trivial task. Its complexity is caused by many factors, such as large spectral or spatial variability as well as the curse of dimensionality. The latter describes the problem of data sparseness. In this particular ill-posed problem, a reliable classification approach requires appropriate modeling of the classification process. The proposed approach is based on a hierarchical clustering algorithm in order to deal with remote sensing data in high-dimensional space. Indeed, one obvious method to perform dimensionality reduction is to use the independent component analysis process as a preprocessing step. The first particularity of our method is the special structure of its cluster tree. Most of the hierarchical algorithms associate leaves to individual clusters, and start from a large number of individual classes equal to the number of pixels; however, in our approach, leaves are associated with the most relevant sources which are represented according to mutually independent axes to specifically represent some land covers associated with a limited number of clusters. These sources contribute to the refinement of the clustering by providing complementary rather than redundant information. The second particularity of our approach is that at each level of the cluster tree, we combine both a high-level divisive clustering and a low-level agglomerative clustering. This approach reduces the computational cost since the high-level divisive clustering is controlled by a simple Boolean operator, and optimizes the clustering results since the low-level agglomerative clustering is guided by the most relevant independent sources. Then at each new step we obtain a new finer partition that will participate in the clustering process to enhance semantic capabilities and give good identification rates.

High and low-level hierarchical classification as an efficient analysis of remotely sensed hyperpectral data

In this paper we address the problem of the remote sensing data analysis on high-dimensional space like hyperspectral data. Its complexity is caused by many factors, such as the large spectral or spatial variability and the curse of dimensionality. Much work has been carried out in the literature to overcome this particular ill-posed issue. Applied to hyperspectral data in hyper-dimensional features spaces, the first particularity of our method is the special structure of its cluster tree. Their leaves are associated with sources represented according to mutually independent axes to represent specifically some land covers. The second particularity is that at each level of the cluster tree we combine a divisive clustering approach applied at a high-level and an agglomerative clustering operating at a low-level. We propose to compare the performances of our approach with those of hierarchical method based on Ward approach.

Nonlinear separation source and parameterized feature fusion for satelite image patch exemplars

We present a new approach for remote sensing image classification. The methodology combines many related tasks namely non linear source separation, feature extraction, feature fusion and learning classification. Nonlinear source separation is a pre-processing stage that aims to compensate the nonlinear mixing natural phenomenon. Latent signals, called sources are transformed to the feature presentation in the feature extraction stage. Feature information presentation is preliminary in machine learning or machine vision projects and provides an efficient and reliable data presentation than original data. Fusing feature aims to enrich the information characteristics about the land cover namely textural information, contours and multi-resolution information. Parameterized fusion model aim to determine the best feature weights in terms of data classification. Finally, a machine learning classification method is used for remote sensing data base. Experimental results show that the proposed fusion method enhances the classification accuracy and provide powerful tool for image exemplars classification.