ABSTRACT. Ultrasound modality has experienced in recent years a paradigm shift not only in technological terms but also in terms of data processing and in clinical applications for both diagnosis and therapy.

The recent introduction of plane wave ultrasound imaging has enabled the development and validation of new imaging modes such as ultrafast Doppler, elastography, super resolution imaging with contrast agent or functional imaging. Thanks to sophisticated signal processing techniques (SVD, artificial intelligence, power processing, advanced engineering, etc.), ultrasound imaging modality offers today superior image quality capable of competing with well-established and very expensive modalities. Beyond the diagnostic dimension, innovative therapeutic applications have emerged to deliver drugs locally or to modulate brain activity.

In this presentation, we will discuss recent developments in ultrasound imaging and present examples of clinical and preclinical applications. Also, we will address the recent applications of ultrasound to treat various diseases including cancer, neurodegenerative diseases or psychiatric disorders.

ABSTRACT. In this study, we have investigated the k-nearest neighbors’ classifier based on magnetic resonance image to perform the diagnostic of pathology on the brain tumor. The system is implemented and simulated in Anaconda, and its performance is tested on the real dataset that contains 6 features and two (02) classes. Each class, abnormal class (patients having brain tumor) and normal class (patients without brain tumor) consists of 2764 instances(images), and 500 instances (images), respectively. For quality analysis, the performance metrics are examined and evaluated. A comparison of the performance characteristics under fixed test sizes (10%~50%) at constant random state 66 is carried out to predict the class status label when k changes from 1 to 20. The simulation results achieved (93.2±6.8)% train accuracy,( 87.0±3.2)% test accuracy, (66.3±17.6)% precision score, (42.6±25.8)% recall, (47.7+21.5)% F1-score , and (95.1±3.1.0) specificity. The proposed method shows the effectiveness of identifying normal and abnormal tumor tissues from the statistical proprieties of the brain MRI image. The performance can be improved by adjusting k and the test size in the range of [2~10],and [15%~25%], respectively. This technique is faster with reduced diagnostic time and effective decisions, serve as a performance metrics guide line that tells us how much better the proposed model is than making predictions. It also helps programmers to select the optimal model that generates the best performance scenario to design a graphical user interface.

ABSTRACT. In this investigation, with primarily studies analysis of the performance metric characteristics we have developed a graphical user interface application to perform the diagnostic of pathology on Parkinson’s disease based on the K-Nearest Neighbor (CKNN) classifier. The system is implemented and simulated in Anaconda, and its performance is tested on real dataset that contains 22 features, with the total of 195 instances and two (02) classes. Each class, abnormal and normal class consists of 48 instances, and 147 instances, respectively. The simulation results achieved better training accuracy, slightly higher than the test accuracy in the range of [75.7%~100%] and [75.0%~92.3%], respectively. The performance prediction of the class status maybe optimized by combining the dataset set size with the k-neighbors parameters. It can be improved by adjusting k and the size in the range of [1~9], and [15%~35%], which results in increasing the interval of the accuracy in the range of [86.7 100 ]% for the training set and [86.4 91,8 ]% for the test set. The proposed method shows the effectiveness of identifying normal and abnormal tumor tissues from the statistical proprieties of the brain MRI image. For quality analysis, the present technique can serve as a test platform for measurement and verification of the pathology disease performance prediction and can be used as a performance metrics guideline to examine the performance that tells us how much better the proposed model is than making prediction. The results maybe will help to design better GUI maximizing the accuracy prediction. helping the medical doctors to diagnostic a patient effectively in a reduced time lapse taking a rapid decision.

ABSTRACT. For the breast tumors classification purpose from ultrasound images, the procedure implemented in this work consists of four steps, namely preprocessing, segmentation, feature extraction and classification. To improve the quality of ultrasound images, anisotropic filtering and histogram equalization are performed on the images, which is the preprocessing step. The segmentation is performed on the image using the Level Set method that allows to extract the region of interest (ROI) and to reduce its size at the same time. Two feature extraction methods are used in this work namely the local binary pattern (LBP) method and the histogram of oriented gradients (HOG) method. The two methods (LBP and HOG) are techniques of textures analysis and allow to characterize the ROI. The extracted feature sets constitute the inputs for three classifiers namely support vector machines (SVM), k-nearest neighbors (KNN) and decision trees (DT). In this work, the best results are obtained by the concatenation of the two feature vectors namely LBP and HOG and its association with the SVM classifier. This allows to achieve an accuracy of 96%, a sensitivity of 97% and a specificity of 94%.

ABSTRACT. Ship detection in Multi-Target Situations has become one of the crucial tasks in maritime surveillance. However, due to the existence of multiple Ship targets with different sizes in seacoast, standard Constant False Alarm Rate (CFAR) detection with a fixed guard window suffers from interfering targets within the training window. As a result, the probability of detection drops drastically. In this paper, a modified fast CFAR detection algorithm based on target indexing in multi-target situations is proposed. The detector does not require any guard window to prevent target interferences in the training window. It consists only of a training window and a test cell. It uses a novel interfering target indexing matrix based on a maximally stable extremal region (MSER) detector that provides the training window with the interference pixels locations to be censored. The Generalized Gamma distribution (GΓD) is adopted as the statistical model of sea clutter. Experimental results show that the proposed method could achieve effective detection results of ships in multi-target situations compared to CFAR detectors with fixed guard window size.

ABSTRACT. Non-orthogonal multiple access (NOMA) is one of advanced technologies which is used to improve the quality of service in fifth generation system. Furthermore, it increases the capacity, massive connectivity, enhanced data rate, reduced latency, and minimizing the required spectrum bandwidth. In the other hand, NOMA is characterized by some disadvantages specially the problem of consuming high power allocation from the users. For this reason, the community of the researchers in wireless communication domain has been developed two techniques such as: the fix and the fair power allocation. In this paper, we compare the sum rate of the fair Power Allocation with the fixed Power allocation algorithm in downlink NOMA systems. Preliminary simulation results have been achieved to confirm that the dynamic fairness power allocation can be useful to perform high sum rate than the fixed power allocation.

ABSTRACT. Breast cancer is the most common cancer among women worldwide. Mammography has become a valuable tool for detecting breast cancer. We were able to achieve the purpose of our study, which was to create an accurate Convolutional Neural Network (CNN) model that classifies mammography images into normal and abnormal using deep transfer learning and data augmentation approaches, to avoid overfitting issues with images. The Mammographic Image Analyses Society MiniMammographic Database (MiniMIAS) was used to train and test the CNN model. The shortage of abnormal images in the MiniMIAS caused a low accuracy, because of that 92 abnormal images were added from the Digital Database for Screening Mammography (DDSM), which leads us to great accuracy. The proposed method starts with a pre-processing step which includes several filters that eliminate any noises and enhance the images, and data augmentation for better training. AlexNet was modified and trained after splitting the dataset into 75%, 5%, and 20% as training, validation, and testing sets respectively. The evaluation results were not very much satisfying with the MiniMIAS database with 96.87%. On the other hand with balanced data, the obtained results were very satisfying with 100%, and better results according to the existing models, which prove that the chosen filters in the preprocessing, as well as the chosen pre-trained model AlexNet, is very useful for breast tumor detection case.

ABSTRACT. Cluster analysis, generally known as clustering, is a technique for separating data into groups (clusters) of similar objects. Except if the system is completely retrained, traditional clustering classifiers will be unable to learn new information and knowledge (attributes, examples, or classes). Only incremental learning, which outperforms when new data objects are introduced into an existing database, can solve this problem. These evolutionary strategies are applied to dynamic databases by updating the data. We’ll choose to study the Incremental Density- Based Spatial Clustering of Applications with Noise algorithm because of its capacity to discover arbitrary clusters and identify noise. In this study, a modified version of the Incremental Density Based Clustering Algorithm using an Adaptive Median Filtering Technique was used. The difference between our previous proposed AMF-IDBSCAN and the proposed algorithm developed in this work is in the evaluation performance stage. The key idea consists of a database change in the case of introducing new data items to an existing database in order to improve performance. We conducted several experiments on benchmark and synthetic data collected from the University of California Irvine repository in terms of the Generalized Dunn Index, Davies Bouldin Index, and change of time (milliseconds) with the increment of data in the original database. Experiments with datasets of various sizes and dimensions show that the proposed algorithm enhances clustering when compared to several current incremental wellknown techniques.

ABSTRACT. Transceiver hardware imperfections impose enormous challenges on the future network deployment, owing to the degraded performance of channel estimation. In this paper, we investigate the impact of the transceiver hardware impairments on the millimiter-Wave (mm-Wave) Multiple-Input Multiple- Output (MIMO) underground channel estimation, in which hybrid beamforming architecture is employed at the transceiver. To accuratley estimate the MIMO underground channel, the compressive sensing theory based multi-resolution hierarchical codebook design is used for the training beamforming. The proposed system operates under the real MIMO underground channel measurements carried out in a mine in Northern Canada. The results show that how hardware impairments greatly degrade the affect performance of the system.

ABSTRACT. From a single human face image, we can collect a wide variety of information, including age, identity, gender and ethnicity. In recent years, age estimation from face image has gained a considerable attention in wide range of applications, various approaches, databases, and evaluation protocols have been proposed. In this paper, the aim of our work is to determine the accuracy of Principal Component Analysis (PCA) method regardless of the influence of external factors after features extraction, and, second, to compare age estimation with and without using dimensionality reduction. We also propose to use Multi-Block Local Binary Pattern (MB-LBP) rather than using Local Phase Quantization (LPQ) and Binarized Statistical Image Features (BSIF) descriptors. To estimate the age of different human face image, we use Support Vector machine Regression (SVR). Our results comparing different of the most commonly Criteria namely Mean Absolute Error (MAE) and Cumulative Score (CS). We prove our methods on FG-NET database and achieve state-of-the-art results for age estimation.

ABSTRACT. Lower limb amputees suffer from mobility limitations that lead to a degradation of their quality of life. Recently, inertial measurement units have been widely used for the analysis and recognition of gait activities which are two essential works for the control of lower limb prostheses. In this paper, accelerometer signals from a dataset are used to recognize gait activities. Five lower limb movements are classified and recognized using linear discriminant analysis (LDA) for dimension reduction and support vector machine (SVM) based on a linear function kernel for classification, respectively. The results prove that classification based on time-domain features and accelerometer signals can achieve 95.69% recognition accuracy for gait activity recognition.

ABSTRACT. Academic institutions define plagiarism as an act of cheating and stealing other’s ideas to pass as their own. Therefore, a huge interest is conducted into plagiarism detection field using multiple techniques. In this article, we propose a method to automatically detect different types of plagiarism from two languages. This method is based on sentence modelling to try to extract plagiarized parts from documents using Doc2Vec model which predicts semantic similarity between documents and phrases.We use the PAN corpus for English plagiarism detection and AraPlagDet for Arabic. Both PAN and AraPlagDet corporas provide a set of suspicious documents that are manually and artificially plagiarized along with their sources.

ABSTRACT. The photometric stereo is a basic domain in computer vision. It consists to reconstruct a 3D object using several intensity images taken by the same camera position and obtained under different lighting conditions. The photometric stereo allows to resolve the ambiguity of the Shape From Shading which reconstruct the 3D object using only one gray scale image and based on the same concepts of the image formation field. The Shape from Shading is considered as an ill-posed problem due to lack of information.

In the literature we can find several works to solve the stereo photometry. In this work we are interested to improve the 3D reconstruction by the mathematical solution "SVD decomposition" which is widely used and very known. In this work we have proposed a new approach with several configurations to solving the Photometric Stereo (PS) using the Particle Swarm Optimization (PSO) meta-heuristic. We also proposed an improvement using the integrability constraint on the 3D surface. We will show that the proposed method gives more optimal and more robust results.

ABSTRACT. Due to the sheer volume of opinion-rich web resources, a lot of current research is focusing on sentiment analysis. The goal is to create new models able to recognize and classify the opinions or sentiments expressed in an electronic text to evaluate and improve a given system. This problem has many solutions proposed in the literature, and most of them are based on Machine Learning techniques. Currently, Deep Learning is a rapidly evolving field, it has proven its effectiveness in many complex problems from its ability to learn and extract meaningful information from data. Therefore, many contributions attempt to adopt this approach in the sentiment analysis area such as that of sentiment classification. In this work, we address the problem of user sentiment analysis for which we propose a new solution based on natural language processing and a deep convolutional neural network. The proposed system is tested on the IMDB dataset that contains 50,000 movie reviews. The obtained results are very persuasive with an accuracy of 99% in the training phase and 89% in the testing phase.

ABSTRACT. this paper presents a system for recognizing the isolated printed Arabic characters. In this type of systems, the feature extraction step is considered as the most important stage to better recognize the characters. For this purpose, we have proposed in this work to use a combined feature methods in order to improve the character recognition rate. The proposed system is based on the combined features extraction, which are: the Discrete Cosine Transform DCT, Zernike moments and the Weighted Gradient Local Binary Pattern WGLBP and the Deep Neural Network (DNN) classifier. Experiments are carried out on the Printed Arabic Text set PAT-A01.The result confirm that the use of the features combination improves the recognition accuracy rate.

ABSTRACT. An elderly person who lives alone is exposed to dangerous situations, for example, a serious fall, a stroke, a heart attack, a serious illness. In these dangerous situations this person risks dying or having serious complications. Among these dangerous situations, we have chosen to study the serious falls of a person, which prevents him from asking for help. In this article we propose a system for detecting serious falls of a person living alone in, intelligent building in an efficient and reliable way, with a minimum rate of false alarms, to ensure instant intervention and reduce risks of injury. This detection is done in video scenes. The serious fall information as well as the images of the event will be sent to another system which informs an assistance center to call for help for this person. This detection is carried out by the classification of the images obtained by the video scenes, which shows the state of this person in each place of the intelligent building. We used the SIFT method and for the parametrization of the images, and in order to obtain a robust system, we opted for a hybrid Neuro-Markovian system, neural networks for the learning phase, and hidden Markov model for the classification phase. In order to validate our obtained results, we carried out a comparative study.

ABSTRACT. In this work, we propose a novel method for ultrasound medical image deconvolution with regularization parameter estimation based on l2 regularization approach and artificial bee colony optimization algorithm. We investigate the impact of regularization choice on the deconvolution process and we evaluate the regularization parameter estimation using our proposed approach. The experimental results confirms the efficient performance of the ultrasound medical image deconvolution operation using our proposed method in comparison with other experimental state-of-the-art deconvolution methods

ABSTRACT. From late November 2019 until today, the Coronavirus (COVID19) has been circulating worldwide on a broad scale. The World Health Organization agrees that this is the world’s most serious outbreak in the last twenty years. Algeria is no exception; although the relatively early precautions taken by the Algerian government, especially the national lockdown, the spread of Coronavirus has attained all cities. This study aims at clarifying how COVID-19 has spread in Algeria so fast in a short time and perform an intelligent system to predict confirmed, deaths, recovered, and active cases. This paper presents firstly a comprehensive analytics study on the spread of the pandemic to understand the COVID19 evolution in Algeria. Secondly, a predictive model based on a Long Short Term Memory Network is proposed and studied to estimate the future number of confirmed, deaths, recovered, and Active cases.

ABSTRACT. Cognitive radio (CR) technology offers the ability to adapt the transmission or reception parameters, thus efficiently improving the wireless spectrum resources utilization. As one of the important components of CR, automatic modulation recognition (AMR) aims to improve the overall communication performance besides promoting efficient and secure transmissions. In this paper, we tackle the semi blind modulation recognition problem in a distributed space-time block coding (D-STBC) scheme for a cooperative relaying network of three nodes, employing the Amplify-and-Forward (AaF) protocol. As classifier, we design a multi-layer artificial neural network (ANN) and train it using higher-order statistics (HOS) as features of the received signals. We evaluate two D-STBC configurations identifying the modulation with and without channel state information (CSI) by exploiting both the temporal and spatial signal dimensions. Experimental results demonstrate that the proposed identifier ensures a good probability for achieving correct modulation identification in acceptable signal-to-noise ratio (SNR) ranges.

ABSTRACT. Subsurface rendering has become essential for photo-realistic rendering; therefore, an accurate simulation is too costly due to the interactions between the particles composing the material and the amount of light passing through it. In this paper, we propose a method to estimate the subsurface light scattering effects. Our method comprises two steps: the first step consists of generating synthetic images using a reference technique to train the neural network. The latter is based on convolutional architecture. The neural network is fed in input by images retrieved from the G-buffer, while in output, it generates an approximation of the subsurface diffusion. The second step consists in generating an approximation of the reference image after a learning step. We show that our method produces effective approximations of subsurface scattering. Furthermore, our method generates images comparable to the reference method and does not require image pre-processing. Therefore, it does not do any calculation in the three-dimensional space of the 3D scene.

ABSTRACT. Gesture Recognition is a technology that makes devices such as a computer capable of recognizing and responding to different gestures produced by the human body. Nonetheless, several factors such as differences in illumination intensity, the complexity of hand gesture models, and other factors can affect the performance of recognition and classification algorithms. Some advances in deep learning such as CapsNets have been suggested to improve the performance of image recognition systems in this particular field. CapsNets emerged to solve part of the limitations of CNN. For this reason, in this work, CapsNets is proposed to solve the American Sign Language problem very effectively. The obtained results showed that the proposed model with a simple data augmentation process produces an test accuracy of 99.08%.

ABSTRACT. In this paper, we present a way to integrate the RTLSDR dongle into the Web. The aim of this paper is to describe a fully functional low-cost embedded system for frequencies analysis based on Raspberry Pi and Realtek Software Defined Radio (RTL-SDR) and using only a web browser to display the spectrum of the received signals, which can be used effectively as an alternative to expensive spectrum analyzer. This proposed method paves the way to Web of Things (WoT) that seamlessly integrates conceptual and physical resources.

ABSTRACT. One of the most important current and emerging challenges for petroleum engineers is to improve the techniques of reservoir characterization in order to reduce the volume of irreducible hydrocarbons. In this regard, accurate determination of attributes related pore-groove geometry and fluid distribution is critical to these reservoir characterization techniques. The most consistent technique that has been developed to identify and classify units with similar pore-groove geometrical attributes (hydraulic units) is called rock typing process. This paper aims at suggesting machine learning based-techniques for reservoir characterization using rock-typing. Both supervised and unsupervised (machine learning) methods have been considered. Applications have been conducted on well log data recorded from ten (10) boreholes drilled in different locations. The obtained results show the efficiency of the proposed approaches, compared with the conventional methods.

ABSTRACT. All radiologists recognize the difficulty of the mammographic examination which requires great expertise and extreme attention. The extension of screening to a larger population, the decrease in the number of specialist radiologists and the quality requirements in terms of public health make it essential to use technologies capable of helping radiologists to treat more cases while improving their performances. Computer-Aided Diagnosis (CAD) meets this need. When you hear about the scientific barriers that have been broken down through the application of deep learning techniques,Convolutional Neural Networks are involved. Also called ConvNets, they are the spearheads of Deep Learning. A CNN is a type of neural network widely used for image classification or visual recognition. The main purpose of a CNN is to extract characteristics from a given image. In this work, Pre-trained models of CNN on ImageNet were exploited and improved to design a perfect system with a lower error rate. The suggested system is essentially composed of two modules: (i) CADe for the detection of anomalies in mammogram that classifies images into normal or abnormal case. (ii) CADx for the identification that classifies the abnormal (cancerous) case as malignant or as benign. The proposed system has been tested on two augmented datasets taken respectively from Digital Database for Screening Mammography (DDSM) and Invasive Ductal Carcinoma (IDC) databases.

ABSTRACT. Wyner-Ziv coding is a particular case of distributed video coding (DVC). It has a low complexity encoding that makes it a good candidate for many wireless applications. One of these applications is in coding the images of the capsule endoscopy, which is a medical technology used for imaging the gastrointestinal tract. One of the problems of DVC is the use of feedback channel (FC) to control the rate, but it can cause a large delay and can increase the decoder complexity. In this paper, we propose to use deep learning to predict the decoding rate and thus to reduce the use of the FC. The deep learning approaches are highly accurate in prediction and can surpass the traditional forecasting approaches. Therefore, a long short-term memory (LSTM) model is used, in our DVC system, to solve the feedback request problem in coding the capsule endoscopy images. The proposed model is implemented in the decoder side to keep the encoder simple. The obtained experimental results are very good and show that the number of requests are significantly reduced and the rate-distortion performance is very near-to-optimal rate and better than those of JPEG.

ABSTRACT. The detection and localization of ships is important and must be accurate and rapid. The Synthetic Aperture Radar (SAR) is optimal for ship detection. Generally, Constant False Alarm Rate (CFAR) algorithms are used to detect ships from a SAR image on the base of satellite remote sensing imaging. However, due to the rapid development of technology, the remote sensing data have shown the features of big-data. The analyze of big-data improves the accuracy and speed of the ship detection. Therefore, Deep Learning (DL) is recommended and exactly the Convolutional Neural Network (CNN) model has greatly improved the static image recognition performance. In this paper, to improve the accuracy and speed of the ship detection from SAR image, we introduce the CFAR-CNN detector. First, after modeling the sea clutter by the Generalized Gamma (GΓ) distribution, a CFAR global detector is applied. Then, to improve the accuracy of the previous results, a CNN local detector is applied. To this end, a real dataset is used to obtain the optimal CNN model. We have shown that this detector is rapid and tends toward ideality.

ABSTRACT. Vehicle detection and tracking is a popular research topic in Intelligent Transportation Systems. The goal of this paper is to detect, identify, and track vehicles in surveillance camera footage in order for them to be extracted efficiently and accurately. You Only Look Once version 4 (YOLOv4) algorithm is used in this paper to propose a real-time vehicle detection and tracking system. The suggested system has been evaluated using a variety of measures, including accuracy, precision, and recognition recall. For the experimental data, the system attained an accuracy of 96.30 percent and an overall accuracy of 94.17 percent. The results reveal that the suggested system successfully tracks vehicles in the scene

ABSTRACT. In this paper, we present a new image encryption approach by generating the key streams using the fractionalorder logistics map, which is due to the advantages it offers such as better chaotic range and more parameters. The algorithm depend on the permutation and diffusion processes, in which pixels are shifted with two step numbers and we also use twovector index to change the locations of columns and rows.To prove the algorithm’s effectiveness in image encryption, it was submitted to a series of analyses and testing, including entropy, histogram, key sensitivity, and other tests.. The simulation results proved that our encryption system has excellent encryption results, in addition to a high security performance compared to various literature. Index Terms—Fractional-order, logistic map

ABSTRACT. The COVID-19 pandemic is an unprecedented global health crisis. Given the delays in obtaining RT-PCR results, the reference diagnostic technique, CT (computed tomography) plays a central role in triaging patients arriving in the emergency department, allowing them to be admitted to "COVID" or "non-COVID" services. To make reliable automatic diagnoses and to quantify the lesion extent, we used deep learning methods for the diagnosis of COVID-19. In this work, we implemented six CNN architectures, namely VGG16, ResNet50, MobileNet, GoogLeNet, Xception, and DenseNet121, and evaluated their performance using appropriate metrics on CT images from freely available public databases. The ResNet50 model was found to give the best results for acceptable computational complexity.

ABSTRACT. Road traffic crashes are among the significant risks facing millions of people around the world every day. Driver fatigue is a salient factor in road accidents. However, overcoming this factor has become possible with the use of artificial intelligence. In fact, with the development of technology, industrial companies in the automotive sector are working on intelligent cars capable of identifying the risks and avoiding them. In this work, we propose a method that identifies driver fatigue. First, we established a comparison between 10 models of Convolutional Neural Networks (CNNs), to classify the state of both mouth and eyes. After selecting the best of them, we compute the percentage of eye closure (PERCLOS) and yawning frequency of mouth (FOM) to conclude the driver state. The proposed method obtained an accuracy of 87.5%.

ABSTRACT. This paper deals with supervised classification of remotely sensed images task and the classifier algorithms used for providing a land cover and land use (LCLU) mapping. Our main contribution is to present two optimization algorithms which are the genetic algorithm (GA) and the artificial immune system (AIS) as a very good interesting alternative to the current traditional classifiers which are for the most part statistical such as a maximum likelihood (MLLH) and Support Vector Machines (SVM) algorithms. Recall that GA and AIS are both bio-inspired methods and belong to the commonly named metaheuristic approach. The most motivations to explore this approach for data classification are also discussed. MLLH, SVM, GA and AIS are applied to classify a multispectral image acquired on June 2001 by ETM+ sensor of Landsat-7 satellite. This multi-band image covers a north-eastern part of Algiers (Algeria). According to the obtained results, two conclusions are reached: firstly, the metaheuristic classifiers are more efficient and robust than the statistical classifiers, and secondly, the AIS classifier may present a promising alternative to the GA classifier for satellite images.

ABSTRACT. Alzheimer's disease (AD) is a neurodegenerative illness that causes structural changes in the brain over time. Furthermore, various symptoms, most notably memory disorder, relate to these structural alterations in the cerebral cortex. Moreover, capturing these changes in the early stages of the disease visually is a difficult task. As a result, creating an algorithm to detect AD disease will be an enormous accommodation for clinicians. The use of machine learning and deep learning-based approaches requires a huge amount of data, which is not the case with medical images. Therefore, we investigated two transfer learning architectures, the VGG16 and the Inception V3, to extract deep features. Following that, principal component analysis was used to reduce vector dimensionality. Three machine learning algorithms were used to perform classification: the support vector machine, adaBoost, and random forest. The proposed investigations have been tested on the Kaggle and ADNI databases. The highest accuracy rates achieved by the Inception V3 feature, and the Random Forest classifier are for the Kaggle database 73.4% and 77.0% for the ADNI database. The proposed model also outperforms the existing state-of-the-art methods.

ABSTRACT. In this article, a new star detection algorithm has been proposed, the algorithm combines two star detection techniques: star detection by the thresholding technique and star detection by filter technique. This method is designed to be tested and used for a star tracker. The proposed method was tested on a set of simulated stellar images (simulated image database) with several types of noises and scenarios encountered in space. The results obtained show the star detection capability of the proposed algorithm

ABSTRACT. The increase of many modern wireless devices in small cells has greatly raised the data traffic overhead. The new 5G technologies are being leveraged to support the massive mobile data traffic. However, the growth in the number of users and their demands can exponentially increase and result in spectrum resource congestion. The spectrum scarcity problem may be solved by dynamically dividing the Millimeter Wave (or Mm-Wave) spectrum for maximizing the utilization of spectrum resources using Cognitive Radio (or CR) technology. In this regard, 5G is not standalone technology innovation but combined with CR technology, it can achieve breakthrough improvements in performance, i.e., compared to previous technologies, namely spectrum utilization efficiency, faster data throughput, and enhanced quality of communication services. This article offers a brief background on 5G technology, how CR works, and how they are combined. In addition, the most widely used mm-wave propagation models have also been presented.

ABSTRACT. In this paper, A mathematical model was developed to study via simulation the performance of a two-stage erbium-doped fiber amplifier (EDFA) model in order to investigate their efficiency when it is implemented in an optical chain, to obtain a flat gain for 8 WDM channels. The EDFA model contains two identical erbium-doped fibers (EDF1, EDF2) separated by an inverted trapezoidal optical filter (ITOF), some other components are used to enhance the signal amplification quality such as an isolator and ASE filter. A simplified dynamic numerical model of EDFA based on a homogeneously broadened two-level approximation of an erbium ion which includes the signal, pump, and both forward and backward ASE powers propagation is used. A uniform spectral shape of gain is obtained for both used configurations, the single EDFA simulation, and the EDFA cascades simulation. The results show the possibility to obtain an equalized output power in good agreement with results in the literature.

ABSTRACT. Future wireless communications systems should be able to accommodate the large number of users and fulfill their diverse quality-of-service requirements. It is widely reported that current wireless systems cannot afford the expected high demands in the data rates. Therefore, innovative physical-layer solutions should be proposed to overcome such a challenge. In this paper, an integrated system architecture for multi-user scenarios is presented. The proposed architecture combines promising techniques to leverage their outstanding overall performance. Specifically, the proposed architecture combines code division multiple access, space time block coding based orthogonal frequency division multiplexing and spatial modulation in an integrated system whose overall performance is shown to outperform state-of- the-art alternatives.

ABSTRACT. This research work presents an accurate method of ‎dimensioning geometric parameters of a microwave multi-‎ports filter design by 3D-HFSS (High Frequency Structure ‎Simulator) simulations based on admittance (Y) matrix ‎extraction technique. The filter's resonant frequency, coupling ‎coefficients and external quality factor can be directly tuned ‎via the multiports Y matrix in a wayit achieves the targetfilter ‎synthesis datasheet set before. A four pole Chebyshev coaxial ‎bandpass filter design is developed showing a good ‎electromagnetic response that presents the availability of the ‎adjustment technique which can be easily adapted to other ‎filter structures.‎

ABSTRACT. The role of Micro-Electromechanical Systems ‎‎(MEMS) ‎antennas in Terahertz (THz) areas has recently graduated due ‎to the ‎modern Computer Aided Design (CAD) techniques. The ‎very special demand for novel MEMS ‎antennas with high ‎performance for wide THz ‎frequency range grows for ‎the ‎foreseeable future, ‎and ‎requires effective optimization ‎processes ‎of improving already proposed designs. This paper ‎describes for the THz wireless access systems a new design of a ‎cone-shaped helix antenna based on MEMS technology. The ‎antenna is developed using tree-dimensional High Frequency ‎Structure Simulator ‎‎(3D-HFSS) for modal electromagnetic ‎analysis, and accurate automatic strategists (Artificial Neural ‎Network ANN) for active optimization. The optimization ‎strategy aims to vary the antenna geometry and maximize it ‎response for the selective frequency band with a high ‎accuracy ‎from Finite Element Method (FEM). ‎The MEMS helix antenna ‎presents very low return losses of less than -63dB at a ‎resonance frequency equals to 3.59 THz, and less than -‎‎20dB ‎for a wide frequency range from 3.12 to 4 THz. Excellent ‎antenna performance and high structure precision are finally ‎achieved by ‎modifying and rectifying different tunable ‎parameters embedded in silicon platform ‎including helix ‎variables.

ABSTRACT. in this article, a novel C-shaped ultra wideband (UWB) Dielectric Resonator Antenna (DRA) With Dual Band-Notched Characteristics for WLAN/X Bands is presented and studied. The proposed DRA structure consists of E-shaped dielectric resonator fed by a stepped Microstrip feed line, and partial ground plane. To realize dual band-notched characteristics of WLAN (5.1–5.9 GHz) and X-band (8.025–8.4 GHz), a Γ-shaped stub is connected to the feed line and inverted C-shaped band stop filter is placed near the Microstrip line. A comprehensive parametric study is carried out using HFSS software to achieve the optimum antenna performance and optimize the bandwidth of the proposed antenna. From the simulation results, it is found that the proposed antenna structure operates over a frequency range of 3.936 GHz to 15.00 GHz with a fractional bandwidth of 116.85%, with two band rejections in the frequency bands of 4.520–6.1 GHz and 7.20–8.424 GHz, which indicating that the antenna is a good candidate for UWB applications, and able to reduces interferences of WLAN band and X-band. The center frequencies and bandwidths of the two notched bands can be adjusted by altering the parameters of the Γ-shaped stub and C-shaped band stop filter.