Search

Bioelectrical time signals are the signals that can be measured through the electrical potential difference across an organ over the time. Electroencephalography (EEG) signals and Electromyography (EMG) signals are among the best-known bioelectrical signals used for medical diagnosis and motion classification. In traditional machine learning methods, the task of extracting unique patterns and features from both bioelectrical signals is hard and requires a specific expert knowledge to study the non-linear, non-stationary and complex nature of these signals. With recent advancement in deep learning methods, features can be extracted from raw data without any handcrafted features. In this paper, a new deep learning approach that integrates EEG with EMG signals is proposed to investigate the efficiency of deep learning in hybrid systems with signal fusion and study the effect of hyper parameters tuning to enhance classification accuracy and boost the performance of hand and wrist motion control without manual feature engineering. Three deep learning models including Convolution Neural Networks CNN model, Long Short Term recurrent neural networks model LSTM, and a combined CNN–LSTM model, were proposed for hybrid system for signal classification. Experiments were tested on a dataset of multi-channel EEG signals merged with multi-channel sEMG signals, to decode hand and wrist motion. Experimental results signify that the proposed deep learning models achieve high classification accuracies that outperform other traditional machine learning state-of-the-art methods with up to 3.5% improvement ratio which indicates the promising application of the approach. Consequently, this work contributes to an automatic classification that facilitates and improves the real-time control of bio-robotics applications, mainly for limb movement classification.

Different respiratory infections cause abnormal symptoms in lung parenchyma that show in chest computed tomography. Since December 2019, the SARS-COV-2 virus, which is the causative agent of COVID-19, has invaded the world causing high numbers of infections and deaths. The infection with SARS-COV-2 virus shows an abnormality in lung parenchyma that can be effectively detected using Computed Tomography (CT) imaging. In this paper, a novel computer aided framework (COV-CAF) is proposed for classifying the severity degree of the infection from 3D Chest Volumes. COV-CAF fuses traditional and deep learning approaches. The proposed COV-CAF consists of two phases: the preparatory phase and the feature analysis and classification phase. The preparatory phase handles 3D-CT volumes and presents an effective cut choice strategy for choosing informative CT slices. The feature analysis and classification phase incorporate fuzzy clustering for automatic Region of Interest (RoI) segmentation and feature fusion. In feature fusion, automatic features are extracted from a newly introduced Convolution Neural Network (Norm-VGG16) and are fused with spatial hand-crafted features extracted from segmented RoI. Experiments are conducted on MosMedData: Chest CT Scans with COVID-19 Related Findings with COVID-19 severity classes and SARS-COV-2 CT-Scan benchmark datasets. The proposed COV-CAF achieved remarkable results on both datasets. On MosMedData dataset, it achieved an overall accuracy of 97.76% and average sensitivity of 96.73%, while on SARS-COV-2 CT-Scan dataset it achieves an overall accuracy and sensitivity 97.59% and 98.41% respectively.

Test cases prioritization has been excessively considered for continious regression and integration testing in Internet of Things based systems to apply multilevel testing activities. Various number of devices, sensors and acctuators are connected together through the internet using different technologies, which requires extensively testing the effeciency of these components and the transferred data between them. Due to the number of the connected components has dramatically increased, causing a direct proportional increase in the number of test cases.Studies that handle the augmentation of the number of test cases for traditional systems lack effeciency when applied for Internet of Things based systems. In this paper, we introduce an enhancement for test cases prioritization using Hill Climbing algorithm as a local search based technique, adapted to achieve tangible effeciency. It is integrated with the LSTM deep learning algorithm for test cases classification purposes. The results of the test cases prioritization using Hill Climbing for regression and integration testing are evaluated using precision, where it achieved 80% and 97% for regression testing, and 93% and 88% for integration testing using two Internet of Things-based system datasets.

Many studies have addressed e-learning, aiming to create a platform for the learning process that completes the traditional classroom work and maximizes the effectiveness of learning outcomes. Gamifying personalized adaptable educational systems have been recently considered to keep the learners motivated and positively progressing in a flow state. However, the current models remain inadequate, providing limited resources for comprehensive learning analytics. In this paper, a theoritical learning preferences adaptation model is proposed based on the Personalized Adaptive Gamified E-learning (PAGE) model. The PAGE model supports blended learning by enforcing the engagement of the traditional learning process’s parties, where effective learning analytics can be sustained to continuously improve the quality of the learning experience. The overall model has been evaluated for its validity through a survey from different perspectives. The overall mean value of the evaluation is 2.77 out of 3. Thus, the evaluation outcomes for the adaptation, gamification, and learning experience of the PAGE model ensure a promising vision for advancements in the learning processes and analytics.

Reducing data stored and transferred is a critical topic in the modern era, particularly after the evolution in multimedia applications and surveillance systems worldwide. Motivated by the massive amount of data generated by surveillance cameras and the enormous number of redundant pixels produced among them, this paper introduces a novel model entitled: “A Smart Multi-View Panoramic Imaging integrating stitching with geometric matrix relations among surveillance cameras (SMPI).” The introduced model aims to create a novel feedback real-time stitching system to reduce the storing and transferring of redundant data generated by neighboring surveillance cameras for an extra level of compression. Moreover, the panoramic view is mostly a better monitoring option rather than multiple monitors in complicated surveillance cameras’ control rooms. The proposed system, in this paper, merges feature extraction stitching techniques with geometric relational matrix calculations to reduce the time complexity limitations of traditional mosaicking. Additionally, the proposed work introduces a real-time algorithm to reconstruct images of each camera from the panoramic view, and a novel algorithm for ordering cameras’ frames before stitching is recommended for producing a panoramic view without any human interference. The experimental work tests numerous state of the art feature extraction algorithms for stitching, Scale Invariant Feature Transform (SIFT), Speed Up Robust Feature (SURF) and Oriented FAST and Rotated BRIEF (ORB) with different orders of stitching. The amount of compression per image after reconstruction is also analyzed. The suggested model was implemented and tested using a vast number of benchmark datasets. Evaluation measures have been used to indicate the efficiency of the recommended system. The proposed model’s algorithm has recorded a low time processing per frame while keeping high accurate results. It was found that the recommended Efficient Stitching Algorithm (ESA) produced an average of 46 panoramas per second, and the reconstruction phase could reach a rate of 90 frames per second, which is significantly higher than the 30 frames per second standard video format system. These results give our model an excellent advantage for the effective processing of more scalable systems with a higher number of frames per second. The proposed system created panoramas with an average of 99% similarities with the traditional mosaicking systems while being highly faster than these conventional methods. Compression ratios and data rate savings, reflecting the gain in data stored and transferred, were calculated, reporting an average of 2.66 and 0.62 per frame, respectively, when applied to standard datasets. The results illustrated that the proposed system gives a dramatic reduction in the volume of data stored/transferred and showed that the creating of mosaics and the reconstruction was made in proper processing time. Experimental outcomes also showed that, for the suggested methods, the produced frames after reconstruction have a high similarity percentage compared with original ones before stitching, which indicates that the proposed approach is efficient enough to preserve the essential features of cameras’ frames without significant information loss.

Tremendous systems are rapidly evolving based on the trendy Internet of Things (IoT) in various domains. Different technologies are used for communication between the massive connected devices through all layers of the IoT system, causing many security and performance issues. Regression and integration testing are considered repeatedly, in which the vast costs and efforts associated with the frequent execution of these inflated test suites hinder the adequate testing of such systems. This necessitates the focus on exploring innovative scalable testing approaches for large test suites in IoT-based systems. In this paper, a scalable framework for continuous integration and regression testing in IoT-based systems (IoT-CIRTF) is proposed, based on IoT-related criteria for test case prioritization and selection. The framework utilizes search-based techniques to provide an optimized prioritized set of test cases to select from. The selection is based on a trained prediction model for IoT standard components using supervised deep learning algorithms to continuously ensure the overall reliability of IoT-based systems. The experiments are held on two GSM datasets. The experimental results achieved prioritization accuracy up to 90% and 92% for regression testing and integration testing respectively. This provides an enhanced and efficient framework for continuous testing of IoT-based systems, as per IoT-related criteria for the prioritization and selection purposes.

During the Coronavirus pandemic, e-learning systems have proven to be an essential pillar for education. This raises to surface what many studies have addressed earlier; creating a platform that completes the traditional classroom work and maximizes the effectiveness of learning outcomes. Striving to achieve such platform, studies have considered gamifying and personalizing the educational resources for the adaptation of educational systems as per the intended learners through intensive learning analytics. But was the learner really a part of the adaptation process taking place? Learning analytics are usually designed to the course’s adaptation and solely for the teachers. Thus, learning analytics in gamified adaptive educational systems involving the course, teachers and learners together are still under investigation. In this study, the Personalized Adaptive Gamified E-learning (PAGE) model is introduced to extend MOOCs by providing new satisfactory levels of learning analytics and visualization in the rich e-learning process that supports the learner’s intervention in the resultant learning analytics. The proposed Learning analytics have been developed to make the necessary adaptation to the course and learner’s learning flow, as well as visualizing the process and adaptation decisions to the learners. Results show a positive potential towards learning adaptation and visualization, and a necessity to provide an additional focus for the gamification concept.

Enormous heterogeneous sensory data are generated in the Internet of Things (IoT) for various applications. These big data are characterized by additional features related to IoT, including trustworthiness, timing and spatial features. This reveals more perspectives to consider while processing, posing vast challenges to traditional data fusion methods at different fusion levels for collection and analysis. In this paper, an IoT-based spatiotemporal data fusion (STDF) approach for low-level data in–data out fusion is proposed for real-time spatial IoT source aggregation. It grants optimum performance through leveraging traditional data fusion methods based on big data analytics while exclusively maintaining the data expiry, trustworthiness and spatial and temporal IoT data perspectives, in addition to the volume and velocity. It applies cluster sampling for data reduction upon data acquisition from all IoT sources. For each source, it utilizes a combination of k-means clustering for spatial analysis and Tiny AGgregation (TAG) for temporal aggregation to maintain spatiotemporal data fusion at the processing server. STDF is validated via a public IoT data stream simulator. The experiments examine diverse IoT processing challenges in different datasets, reducing the data size by 95% and decreasing the processing time by 80%, with an accuracy level up to 90% for the largest used dataset.

the dangers generated from synthesized multimedia are increasing every day. The creation of the so-called Deepfakes multimedia is vastly evolving, making the detection task harder every day. Researchers and corporations are interested in exploring the technology limits and are coming up with new tools every year to create more robust fake media. In this paper, a new enhanced fake video detection model is introduced addressing many of the face-swapping threats and the low generalization problem. A preprocessing stage is proposed to minimize the noise in the data to enhance their quality. The proposed architecture uses a modified application of capsule neural networks (CapsNet) with an enhanced routing technique. It does not require a lot of training data and generates a small number of training parameters making it fast to build. The model was trained and tested using the DFDC-P dataset and the results have proven that it outperformed other detectors in terms of detection recall, weighted precision, and F1 score.

Securing electronic medical information has a significant impact on data access which directly affects patient’s privacy and quality care rights. Medical professionals need to have full access to all patients' medical history to make accurate decisions regarding diagnosis and treatment plans. This paper introduces a novel Secure framework that integrates a new hybrid encryption scheme of medical images using chaotic maps and 2D Discrete wavelet transform (DWT) Steganography to increase key size and achieve high security level. In addition, a web-based monitoring platform has been deployed for tracking of electronic medical records during transmission. To validate the efficiency of the proposed framework, an application case-study has been introduced for securely transmitting retinal fundus medical images for diagnostic decisions on diabetic patients. Compared to the state-of-the-art approaches, the proposed framework demonstrated the ability to mask the context of the confidential patient into a transmitted cover image with high imperceptibility and limited degradation in the stego image provided in acceptable cryptographic overhead. Experimental results demonstrated that the proposed framework outperforms other schemes in terms of accuracy, sensitivity and perceptibility.

Optical Coherence Tomography (OCT) imaging has major advantages in effectively identifying the presence of various ocular pathologies and detecting a wide range of macular diseases. OCT examinations can aid in the detection of many retina disorders in early stages that could not be detected in traditional retina images. In this paper, a new hybrid computer-aided OCT diagnostic system (HyCAD) is proposed for classification of Diabetic Macular Edema (DME), Choroidal Neovascularization (CNV) and drusen disorders, while separating them from Normal OCT images. The proposed HyCAD hybrid learning system integrates the segmentation of Region of Interest (RoI), based on central serious chorioretinopathy (CSC) in Spectral Domain Optical Coherence Tomography (SD-OCT) images, with deep learning architectures for effective diagnosis of retinal disorders. The proposed system assimilates a range of techniques including RoI localization and feature extraction, followed by classification and diagnosis. An efficient feature fusion phase has been introduced for combining the OCT image features, extracted by Deep Convolutional Neural Network (CNN), with the features extracted from the RoI segmentation phase. This fused feature set is used to predict multiclass OCT retina disorders. The proposed segmentation phase of retinal RoI regions adds substantial contribution as it draws attention to the most significant areas that are candidate for diagnosis. A new modified deep learning architecture (Norm-VGG16) is introduced integrating a kernel regularizer. Norm-VGG16 is trained from scratch on a large benchmark dataset and used in RoI localization and segmentation. Various experiments have been carried out to illustrate the performance of the proposed system. Large Dataset of Labeled Optical Coherence Tomography (OCT) v3 benchmark is used to validate the efficiency of the model compared with others in literature. The experimental results show that the proposed model achieves relatively high-performance in terms of accuracy, sensitivity and specificity. An average accuracy, sensitivity and specificity of 98.8%, 99.4% and 98.2% is achieved, respectively. The remarkable performance achieved reflects that the fusion phase can effectively improve the identification ratio of the urgent patients&rsquo, diagnostic images and clinical data. In addition, an outstanding performance is achieved compared to others in literature.

Data streams have become a widely-adopted data representation format in many real-world applications. This data streaming may be published for different scientific research, mining, or analysis purposes. However, such streams may contain personal-specific data that could be considered as sensitive about individuals. This makes the privacy preserving of data streams against privacy disclosure attacks, while maintaining their utilization, is a real challenge. Some studies have considered privacy-preserving publishing over data streams with only Single Sensitive Attribute, in which they do not protect the published streams from all possible privacy attacks. In this paper, we propose a novel Restricted Sensitive Attributes-based Sequential Anonymization (RSA-SA) approach for privacy-preserving data stream publishing. Besides, two new privacy restrictions are introduced to restrict the published Sensitive Attributes values: Semantic-diversity and Sensitivity-diversity. RSA-SA can protect the sensitive values of the published data streams against the related privacy attacks, including the attribute disclosure, skewness, similarity, and sensitivity attacks. In addition, RSA-SA handles data streams that have either single or multiple sensitive attributes with minimum information loss and delay time. Thus, the data utility of the published data streams is efficiently maintained to provide more accurate mining and analytical results, where robust invulnerability to privacy attacks is sustained.

Malware is the umbrella term that denotes attacking any system by malicious software. During the last few years, the popularity of Android smartphones led to the sneak of several malware applications into different Android markets without any difficulty. As a consequence of this, malware applications have been grown exponentially in the Android markets. Unfortunately, most of these markets suffer from an inability to detect malware, which results in increasing the probability of infecting users’ phones with these malware applications. The present paper focuses on developing an efficient computational framework based on Deep Belief Networks for malware detection. The proposed framework merges high level static analysis, dynamic analysis and system calls in feature extraction in order to achieve the highest accuracy. The evaluation compares the most familiar machine learning approaches that were applied in malware detection with the proposed framework. The obtained results demonstrate that Deep Belief Networks technique can realize 99.1% accuracy with the presented dataset. Over and above that, we develop our complete static analysis jar which adopts different efficient methods in an attempt to facilitate and speed up the static analysis by handling all the Android applications in only one step rather than considering one application at a time. Keywords: Android, Static analysis, Malware detection, Dynamic analysis, System calls, Deep Belief Networks, Deep learning

Detecting and classifying cardiac arrhythmias is critical to the diagnosis of patients with cardiac abnormalities. In this paper, a novel approach based on deep learning methodology is proposed for the classification of single-lead electrocardiogram (ECG) signals. We demonstrate the application of the Restricted Boltzmann Machine (RBM) and deep belief networks (DBN) for ECG classification following detection of ventricular and supraventricular heartbeats using single-lead ECG. The effectiveness of this proposed algorithm is illustrated using real ECG signals from the widely-used MIT-BIH database. Simulation results demonstrate that with a suitable choice of parameters, RBM and DBN can achieve high average recognition accuracies of ventricular ectopic beats (93.63%) and of supraventricular ectopic beats (95.57%) at a low sampling rate of 114 Hz. Experimental results indicate that classifiers built into this deep learning-based framework achieved state-of-the art performance models at lower sampling rates and simple features when compared to traditional methods. Further, employing features extracted at a sampling rate of 114 Hz when combined with deep learning provided enough discriminatory power for the classification task. This performance is comparable to that of traditional methods and uses a much lower sampling rate and simpler features. Thus, our proposed deep neural network algorithm demonstrates that deep learning-based methods offer accurate ECG classification and could potentially be extended to other physiological signal classifications, such as those in arterial blood pressure (ABP), nerve conduction (EMG), and heart rate variability (HRV) studies.

The amount of data collected by various organizations about individuals is continuously increasing. This includes diverse data sources often for data of high dimensionality. Most of these data are stored in tabular format and can include sensitive content. Preserving data privacy is an essential task in order to allow such data to be published for different research and analysis purposes. In this context, Privacy-Preserving Tabular Data Publishing (PPTDP) has drawn considerable attention, where different approaches have been proposed to preserve the privacy of individuals' tabular data. Such data can include Single Sensitive Attributes (SSA) or Multiple Sensitive Attributes (MSA) or come from data streams. In this paper, we conduct a comprehensive study to analyze and evaluate the main different data anonymization approaches that have been introduced in PPTDP. The study investigates the three broad areas of research: SSA, MSA and data streams. A detailed criticism is presented to highlight the strengths and the weaknesses of each approach including their deployment in the cloud and Internet of Things (IoT) environments. A research gap analysis is discussed with a focus on capturing current state of the art in this field in order to highlight the future directions that can be considered.

Wireless visual sensor networks (WVSNs) are composed of a large number of visual sensor nodes covering a specific geographical region. This paper addresses the target detection problem within WVSNs where visual sensor nodes are left unattended for long-term deployment. As battery energy is a critical issue it is always challenging to maximize the network’s lifetime. In order to reduce energy consumption, nodes undergo cycles of active-sleep periods that save their battery energy by switching sensor nodes ON and OFF, according to predefined duty cycles. Moreover, adaptive compressive sensing is expected to dynamically reduce the size of transmitted data through the wireless channel, saving communication bandwidth and consequently saving energy. This paper derives for the first time an analytical framework for selecting node’s duty cycles and dynamically choosing the appropriate compression rates for the captured images and videos based on their sparsity nature. This reduces energy waste by reaching the maximum compression rate for each dataset without compromising the probability of detection. Experiments were conducted on different standard datasets resembling different scenes; indoor and outdoor, for single and multiple targets detection. Moreover, datasets were chosen with different sparsity levels to investigate the effect of sparsity on the compression rates. Results showed that by selecting duty cycles and dynamically choosing the appropriate compression rates, the desired performance of detection can be achieved with adaptive CS and at the same time saving energy, where the proposed framework results in an 70% on average energy saving as compared to transmitting the captured image without CS.

Decomposition based Multi Objectives Evolutionary Algorithms (MOEA/D) became one of research focus in the last decade. That is due to its simplicity as well as its effectiveness in solving both constrained and unconstrained problems with different Pareto Front (PF) geometries. This paper presents the challenges on different research areas concerning MOEA/D. Firstly, the original MOEA/D algorithm is explained. Its basic idea is to decompose the Multi Objectives Optimization (MOO) problem into multiple single objective optimization sub problems and works concurrently to solve these sub problems. Each sub problem is optimized with the help of the information gained from its neighborhood. Then, two major factors that affect the search ability of decomposition based MOO algorithms: Scalarization Functions (SF) and weight vectors generation and adaptation are discussed. Furthermore, the researches in two categories of different variants of MOEA/D are illustrated. Finally, the real world application areas that applied the decomposition approach are mentioned.

Recently, service-oriented computing paradigms have become a trending development direction, in which software systems are built using a set of loosely coupled services distributed over multiple locations through a service-oriented architecture. Such systems encounter different challenges, as integration, performance, reliability, availability, etc., which made all associated testing activities to be another major challenge to avoid their faults and system failures. Services are considered the substantial element in service-oriented computing. Thus, the quality of services and the service dependability in a web service composition have become essential to manage faults within these software systems. Many studies addressed web service faults from diverse perspectives. In this paper, a comprehensive study is conducted to investigate the different perspectives to manipulate web service faults, including fault tolerance, fault injection, fault prediction and fault localization. An extensive comparison is provided, highlighting the main research gaps, challenges and limitations of each perspective for web services. An analytical discussion is then followed to suggest future research directions that can be adopted to face such obstacles by improving fault handling capabilities for an efficient testing in service-oriented computing systems.

Internet of Things (IoT) systems are fast evolving nowadays, in which huge amounts of data are produced rapidly from heterogeneous sources. The nature of IoT-based systems implies many challenges, in terms of operation, security, quality control and data management. Thus, testing such systems is a key element to their success. In this paper, we present a comprehensive study for the main testing techniques and tools that have been considered for the IoT-based systems. Detailed comparison and analytical criticism are conducted, identifying the different testing types that have been applied for the main application domains. The research gaps are addressed, which highlight the future directions that can be adopted.

Gesture recognition has been recognized as a natural way for the communication especially for elder or impaired people. Hand gesture recognition is an important research issue in the field of human-computer interaction (HCI), because of its extensive applications in virtual reality, sign language recognition, and computer games. There is a number of algorithms addressing different aspects of the Gesture recognition problem have been proposed. While image-based techniques have been widely studied, it may be affected by lighting conditions, large variations of the hand gesture and textures. Recently, with the developments of new technologies and the large availability of inexpensive depth sensors, real time gesture recognition has been faced by using depth information and avoiding the limitations due to complex background and lighting situations. This paper introduces an enhanced automated model for hand gesture recognition using convolution neural network (CNN). In this paper, a hand gesture recognition model with Kinect sensor has been proposed, which operates robustly in uncontrolled environments and is insensitive to hand variations and distortions. The proposed model uses both the depth and color information from Kinect sensor to detect the hand shape, which ensures the robustness in cluttered environments. The proposed model consists of two major modules, namely, hand detection and gesture recognition. Experiments have been conducted on large dataset to demonstrate the efficiency of the proposed model. The experimental results show an outstanding performance in the terms of accuracy, recall and precision.

Popularity of social networking sites increases day by day. Due to its high popularity, the number of comments can grow at a high rate when a social message published. The users of social networking sites may wish to get a brief knowledge of comment streams without reading the all comments. The system aim to group comments with similar opinion together and create a summary for this message after filtering abusive words from each generated comments. Since the users will request the summary at any moment, we use IncreSTS algorithm for realtime incremental clustering of social network comment streams. Besides, we extract key-terms from each clusters that help users to easily get an overview summary.

A new automated seizure detection model is proposed.Applying Weighted Permutation Entropy to classify raw and decomposed EEG signals.Simulating physiological and environmental artifacts to test the model robustness against noise.Comparison between Support Vector Machine and Artificial Neural Network classifier.Comparison with other automatics seizure detection models found in literature. The paper introduces a new automated seizure detection model that integrates Weighted Permutation Entropy (WPE) and a Support Vector Machine (SVM) classifier model to enhance the sensitivity and precision of the detection process. The proposed system utilizes the fact that entropy based measures for the EEG segments during epileptic seizure are lower than in normal EEG. The new suggested model better tracks abrupt changes in the signal and assigns less complexity to segments that exhibit regularity or are subjected to noise effects. The Weighted Permutation Entropy algorithm relies on the ordinal pattern of the time series along with the amplitudes of its sample points. The proposed technique is implemented and tested on hundreds real EEG signals and the performance is compared based on sensitivity, specificity and accuracy. Various experiments have been applied in different scenarios including healthy with eyes open, healthy with eyes closed, epileptic patients during no-seizure state from two different location of the brain. Other scenarios have been applied accompanied by background simulated noise resulting from physiological and environmental artifacts. Results showed outstanding performance and revealed promising results in terms of discrimination of seizure and seizure-free segments. It also manifests high robustness against noise sources.

Applications for the Internet of Things (IoT) have evolved in excessive numbers, producing a vast amount of data needed for intelligent processing. The varying IoT infrastructures such as cloud and IoT application layer protocol limitations in the transmission/receiving of messages become the barriers in the implementation of intelligent IoT apps. In this paper, we review the importance of Big data, cloud computing and fog computing in IoT and the challenges of using machine learning in IoT. Finally, we discuss the general statistics of using artificial intelligence in IoT applications.

Saving data stored and transferred is a critical topic in modern era, especially after the evolution in the numbers of surveillance cameras in the world. So an extra level of compression is needed to contribute in solving the problem of the huge data produced by those cameras. The proposed system: “An Efficient multi-view panoramic imaging and Extra Compression of surveillance cameras’ footage using Stitching” (ECS) aims to contribute in the solution of this problem by merging feature extraction stitching techniques with geometric relational matrix calculations. ECS creates a new real-time efficient stitching algorithm to minimize the sending and storing of the redundant pixels between the footage of the surveillance cameras. The limitations of using stitching on video footage and not on still images is due to the high complexities of stitching algorithms; the proposed system, in this paper, uses the projective affine transformation matrix between images of camera footage to reduce this limitations. The work presented in this paper introduces, for the first time, an algorithm to reconstruct images of each camera from the panoramic view. The experimental work of this research tests multiple feature extraction algorithms for stitching; Scale Invariant Feature Transform (SIFT), Speed Up Robust Feature (SURF) and Oriented FAST and Rotated BRIEF (ORB) with different order of stitching, and the amount of compression per image after reconstruction is analyzed. Moreover, the panoramic view is mostly a better monitoring option rather than multiple monitors in complex surveillance cameras’ control rooms. Experimental results show that the proposed method could be faster than other state-of-the-art image stitching methods and accurate panoramic image. The system also gives a better visual effect in a large view panorama due to the high speed real time stitching introduced in this paper.

Deep learning algorithms have achieved high performance accuracy in complex domains such as image classification, face recognition sentiment analysis, text classification, and speech understanding. Due to the nested non-linear structure of deep learning algorithms, these highly successful models are usually applied in a black-box manner, i.e., no information is provided about what exactly causes them to arrive at their predictions. The effectiveness of these systems is thus limited by the machine’s current inability to explain its decisions and actions to human users. Such a lack of transparency can be a major drawback. For instance, in medical applications the development of methods for visualizing, explaining, and interpreting deep learning models has recently attracted increasing attention. Explainable Machine Learning (XAI) or Interpretable Machine Learning (IML) programs aim to create a suite of machine learning techniques that produce more explainable models while maintaining the high level of accuracy. It also enables human users to understand, appropriately trust, and effectively manage the emerging generation of artificially intelligent partners. Our work summarizes recent developments in this field, and we present three classification tasks wherein we make use of LIME (Local Interpretable Model-Agnostic Explanations) to explain predictions of deep learning models. LIME attempts to make these complex models at least partly understandable by evaluating using three classification tasks. Our first evaluation is performed on natural language processing application for tweet data classification and our second evaluation is on biomedical signal classification cancer detection. Our third evaluation is on Windows PC malware classification in a cybersecurity domain. In all three case studies, the interpreted results are presented in terms of visualization.

Melanoma is the deadliest form of skin cancer. Unfortunately, Skin cancer can’t be identified by visual examination. So, there is a call for an automated model which assists dermatologists in early diagnosis of skin cancer and help remote patient to save their life by remote diagnosis. This paper introduces an enhanced expert computer-aided model for skin diagnosis using deep learning. The proposed region of interest (ROI) segmentation is done by integrating both color and texture properties for the skin in both spatial and frequency domains. Then, the convolution neural network (CNN) is used for extracting all the possible discriminating features. Experiments have been conducted on various large datasets to demonstrate the efficiency of the proposed model. The experimental results show an outstanding performance in the terms of sensitivity, specificity and accuracy compared with others in literature.

The investigation of clinical reports suggested that more than ten percent patients with diabetes have a high risk of eye issues, since eye is a very sensitive organ in human body; it responsible for sensing information in the visible light domain, So, early detection of any abnormal features in eye fundus retinal is essential. Experts need more professional automatic detection way to detect the yellowish spots caused by diabetes; to control its propagation on the retina. The retinal image analysis is a natural approach to deal with eye diseases and other diseases that affecting eye fundus. The paper introduces a new medical diagnosis model for optic disc detection in fundus images by automated segmentation of landmarks. Automatic segmentation of retinal blood vessels from retinal images is applied. The proposed model integrates adaptive Otsu's Threshold, Hough transform and kirsch's template filtering for automated detection of candidate optic nerves. Traditional segmentation methods which are based on spatial properties of the optic disc achieve good results but usually fail to detect the optic disc in pathological images where abnormalities found, such as large exudates, are confused with the optic disc due to their similar appearance, color, and texture. A proposed segmentation model is introduced, which integrates vessels detection mechanism to achieve better sensitivity rates and high performance. Experimental results showed that the number of false responses are greatly reduced in the presence of other similar abnormal signs found in the eye fundus retina images. Experiments showed that proposed model produced high performance in detection of optic disc, blood vessels and exudates under different infection. More aware, experiments results showed high sensitivity compared with recent researches.

Electroencephalography (EEG) and Electromyography (EMG) signals are playing significant role in controlling bio-robotics applications, such as prostheses. Brain Computer Interfaces (BCIs) for amputees allow them to use their remaining functionalities as control possibilities by turning brain signals into commands for external devices. However, BCIs that use the EEG signals alone are not yet fully acceptable in bio-robotic applications. Myoelectric control systems use EMG signals recorded on the residual muscles of amputated limbs to control prosthesis, but the residual muscles cannot often provide enough signal to control a multiple degrees of freedom prosthetics. This paper introduces a new hybrid BCI model that integrates EEG and EMG signal processing with machine learning models to efficiently improve classification accuracy and increase the control performance of different upper limb movements for above elbow amputees. Experiments have been carried out on a large dataset of 64 channel EEG signals, combined with 32 channel surface EMG (sEMG) signals, acquired simultaneously from above-elbow amputees to decode five hand and wrist motions. Classification accuracy is used for evaluating the classification performance with 5-fold cross validation. Experimental results demonstrate that the proposed model achieves a high classification accuracy, exceeding 98.8%, using 6th order Autoregressive (AR) model coefficients with three proposed combined set of features from time domain, frequency domain and entropies. This suggested model outperforms previously proposed models in the literature with up to 8.9% improvement ratio; this implies that Hybrid-BCI can be reliably used to improve the control performance of upper limb movements. Furthermore, the model may be easily applied into real-time hybrid BCI application.

Wireless Visual Sensor Networks (WVSNs) have gained significant importance in the last few years and have emerged in several distinctive applications. The main aim is to design low power WVSN surveillance application using adaptive Compressive Sensing (CS) which is expected to overcome the WVSN resource constraints such as memory limitation, communication bandwidth and battery constraints. In this paper, an adaptive block CS technique is proposed and implemented to represent the high volume of captured images in a way for energy efficient wireless transmission and minimum storage. Furthermore, to achieve energy-efficient target detection and tracking with high detection reliability and robust tracking, to maximize the lifetime of sensor nodes as they can be left for months without any human interactions. Adaptive CS is expected to dynamically achieve higher compression rates depending on the sparsity nature of different datasets, while only compressing relative blocks in the image that contain the target to be tracked instead of compressing the whole image. Hence, saving power and increasing compression rates. Least mean square adaptive filter is used to predicts target's next location to investigate the effect of CS on the tracking performance. The tracking is achieved in both indoor and outdoor environments for single/multi targets. Results have shown that with adaptive block CS up to 20 % measurements of data are required to be transmitted while preserving the required performance for target detection and tracking.

MicroRNAs are small non-coding RNA molecules that play an important role in post-transcriptional gene regulation. They bind with messenger RNAs, affecting the regulation of gene expression causing diseases such as heart diseases and cancer. The two pioneer strategies of miRNA target prediction are either experimentally or computationally. Since experimental techniques are very expensive and slow, several computational tools have been proposed to defeat some of the experimental technical difficulties. In this paper, we propose an enhanced computational framework that predicts miRNA targets depending on structural, positional, and thermodynamic features extracted from the alignment of miRNA and its targets. Then, we select the linearly separable features using the eigenvalue analysis of the covariance matrix of the features. Finally, the selected features are applied to random forest classifier. The obtained results prove that our framework significantly excels other existing tools.

With the recent remarkable and fast evolution in telecommunication and computing technologies, great amounts of individuals' tabular-formatted data are collected and used by several organizations in the society. In some cases, some organizations need to share these gathered data to be used in business analysis, decision making or scientific researches purposes, which can involve sensitive information about the individuals. However, these data cannot be published in their original form to other third parties due to the associated privacy concerns. Consequently, preserving individuals' privacy represents a critical issue when sharing the individuals' private data. Hence, Privacy-Preserving Tabular Data Publishing (PPTDP) has received a great attention to protect the privacy of individuals' tabular data, where several approaches have been presented to address this issue. In this paper, we propose an enhanced additive noise approach for privacy-preserving microdata with Single Sensitive Attribute (SSA) publishing. The proposed approach maintains better published data utility to allow more accurate mining and analytical results, where more robust privacy protection against privacy attacks is provided.

One of the most challenging problems in the computational biology is inferring the evolutionary history to construct phylogenetic trees. A phylogenetic tree is a branching topology that represents the evolutionary relationships among groups of organisms. Traditional construction methods of phylogenetic trees rely on building a multiple sequence alignment (MSA) of the corresponding sequences of organisms. However, achieving optimal MSA for a numerous number of sequences is computationally complex. Moreover, even if an optimal MSA is obtained, it may not reflect the evolutionary history of the sequences. So, several free-alignment methods for phylogenetic tree construction have been proposed. In this paper, we investigate a free-alignment construction method to infer the optimal phylogenetic tree based on differential evolution algorithm. The evaluation results confirm that the investigated method succeeds in building the optimal tree with accuracy higher than other considered methods.

One of the most important challenges that face researchers in the medical field is the variety and the extraordinary mass of data types that can be generated even for a single patient. Machine learning models are commonly used to extract knowledge from medical data. However, most of these models have been built based on only a single data type, ignoring learning valuable information existing in other data types. In this paper, we seek to enhance the efficiency of machine learning models in knowledge extraction. We apply data integration via ensemble stacked supervised generalization method. We propose a framework which adopts feature selection techniques to improve the stacked models in order to achieve a better performance than state-of-the-art data integration models. We also consider getting information from multiple data sources using multiple machine learning models to explore data effectively as each data type has a suitable machine learning model. The proposed predictive model is applied to the METABRIC dataset with survival prediction of breast cancer at 2000 days of diagnosis. Our ensemble stacked generalization method achieves 80% prediction accuracy which represents a high-performance index relative to the other known ensemble methods. Also, we demonstrate that some enhanced feature selection methods are able to select relevant features that are close to the features selected by a biological expert from breast cancer pathways.

Gamifying the educational systems implies the application of gaming rules in a non-gaming context. This has positively-affected the success and engagement of learners in the learning process. The learning styles work as an important factor for a better definition of the preferred way for a learner to learn, resulting in an adaptable system that understands the educational needs of the learner, with adaptable teaching styles that draw his attraction to continue learning. In this paper, we discuss the state-of-art principles needed for serious gamification, and investigate the main concepts required to create adaptive educational systems. This includes the study of the adaptation types considered for both normal and disabled/disordered learners, as well as building a detailed comparison between the different evaluated educational models, while highlighting the concepts applied and analyzing their strengths and weaknesses.

Wind energy projects have recently been associated with huge investments. This led researchers to dig more into managing the costs of wind farms. The operation and maintenance (O&M) costs have a big effect on the success of wind farm projects. Thus, monitoring wind turbines and predicting the O&M costs become of a crucial demand. O&M costs are related to some main parts in wind turbines, like the spare parts and rotor blades that are exposed to damage because of the surrounding environmental factors. Hence, they can be monitored via deployed sensors that generate massive quantities of incomplete, and heterogeneous data. Therefore, a big data analytical system is required to analyze these data. In this paper, we propose the Wind Farms Economic Classifier (WFEC) System that uses big data analytics to manage the data volume, variety and veracity to predict the O&M costs. WFEC proposes an enhanced Flexible Naive Bayes Classifier (FNBC) to classify wind farms profitability according to the predicted O&M costs. Experiments show that WFEC achieves high classification accuracy with less processing time.

This paper introduces a new model for arm and leg gestures recognition in a video stream. The proposed model recognizes a set of six specific hand gestures and two specific leg gestures, namely: Full Right-Hand Wave, Full Left-Hand Wave, Full left and right wave, Top Right-Hand Wave, Top Left-Hand Wave, Top both hand wave, Right leg, Left leg, Clapping. The proposed model consists of five phases: video acquisition and preprocessing, video segmentation or depth segmentation, detection, tracking, and classification. The proposed model overcomes the limitations found in previously proposed models as it can be applied on non-stationary background, can deal with noisy video input and requires less time consumption. The advantages mentioned above are achieved due to our most important contribution in the selection of the suitable algorithm that performs our goal efficiently in each phase of the proposed model. The proposed model integrates the separation of foreground movements in video segmentation phase with multi-layer Viola-Jones algorithm in detection phase. The output of these two phases is deployed in tracking motion of the moving region of interest using clustering of feature points, the output of which can be used for understanding simultaneously performed body hand-leg gestures. Our framework uses Kinect camera to connect video streams and integrates various techniques to make tracking tasks efficient. Experiments have been carried out to demonstrate the effectiveness of the proposed model on different benchmarked datasets and a newly generated dataset that was made specifically for this proposed model. IXMAS, Weizmann, and G3d Depth dataset have been used to validate the proposed model which demonstrated an outstanding improvement regarding accuracy for the iXMAS dataset and the G3d dataset.

Medical monitoring requires instant visibility across data sources and access to dynamic analyses. However medical monitoring among patients, perform in-stream medical advice, remains a challenging problem. Blood banks suffer frequent shortage of blood due to lack of blood donations, hence blood donation requests are frequently seen on social media for patients who urgently require blood transfusion with specific blood group. Recently, worldwide efforts have been undertaken to utilize social media and smart phone applications to make the blood donation process more convenient and provide a concrete information system that allows donors and blood donation centers to communicate efficiently and coordinate with each other to minimize time and effort required for blood donation process. This paper aims at developing a Cloud medical monitoring and Web-Based Blood Donation System which will allow blood donors and patients to offer/request blood donation from blood banks. Additionally, a new method is proposed for continuous observation and communication among doctors and patients. Using IOT cloud platform, simple medical devices equipped with medical sensors can monitor health status of patients and update the electronic medical records of patients’ information. Medical experts can remotely monitor patients dynamic status and give prompt medical advice. The developed Web-Based application utilizes a cloud based hosting platform to enhance system performance and ensure high availability. A mobile application has been developed where users will be able to use as an application installed on their smart phones to help them complete blood donation process with minimum effort and time. This application helps people receive remotely medical advices and helps establish a blood donation community through social networks. This paper also presents various tools that were used to measure system performance.

An adaptive Fuzzy Inference Perceptual model has been proposed for watermarking of video streams. The model is designed to be adaptive to the human visual characteristics to attest the owner identification and discourage the unauthorized copying. The proposed model integrates the human visual characteristics of video motion sub-regions, in the frequency multi-resolution wavelet domain, with multi-dimensional fuzzy inference perceptual model. The designed fuzzy multi-variable input system inherits luminance, texture and gradient to generate an adaptive watermark embedding strength factor. Interacting motion estimation is applied for the selection of the candidate frames and sub-blocks. A fuzzy-based scheme has been proposed to generate a perceptual membership degree for strength watermark embedding factor for the selected candidate motion blocks. Experiments have been carried out with different benchmark test videos of different sizes, visual characteristics and sampling rates. Various sizes of watermark signatures have been applied on the model. Several experimental attacks have been applied such as frame dropping, frame averaging, JPEG compression, Gaussian noises. It manifests considerable robustness against various geometric and signal processing attacks. Several attacks have been applied to the proposed scheme and the experiments revealed promising results in terms of visual quality and extracted watermark distortion. In addition, the model has been compared with different other watermarking schemes in literature. The proposed model showed superior performance in terms of fidelity, robustness to attacks and high level of imperceptibility.

financial researches showed that technical indicators are useful tools for stock prediction. Technical indicators are used to generate trading signals (buy/sell) signals. The main problem of an indicator usage is to determine its appropriate parameters. In this paper a new GA based technique for optimizing the parameters of a collection of technical indicators over two objective functions Sharpe ratio and annual profit is proposed. The technique handles four indicators DEMAC (Double Exponential Moving Average Crossovers), RSI (Relative Strength Index), MACD (Moving Average Convergence Divergence), and MARSI (Moving Average RSI) indicators. The technique was tested on 30 years of historical data of DJIA (Dow Jones Industrial Average) stock index. Results showed that the optimized parameters obtained by the proposed technique improved the profits obtained by the indicators with their typical parameters, the Buy and Hold strategy and the random strategy.

With recent progress in pervasive healthcare, physical activity recognition with wearable body sensors has become an important and challenging area in both research and industrial communities. Here, we address a novel technique for a sensor platform that performs physical activity recognition by leveraging a class specific regularizer term into the dictionary pair learning objective function. The proposed algorithm jointly learns a synthesis dictionary and an analysis dictionary in order to simultaneously perform signal representation and classification once the time-domain features have been extracted. Specifically, the class specific regularizer term ensures that the sparse codes belonging to the same class will be concentrated thereby proving beneficial for the classification stage. In order to develop a more practical approach, we employ a combination of an alternating direction method of multipliers and a l 1 − l s minimization method to approximately minimize the objective function. We validate the effectiveness of our proposed model by employing it on two activity recognition problem and an intensity estimation problem, both of which include a large number of physical activities. Experimental results demonstrate that classifiers built in this dictionary learning based framework outperforms state of art algorithms by using simple features, thereby achieving competitive results when compared with classical systems built upon features with prior knowledge.

Motor neuron disease (MND) is a neurological disorder that attacks and destroys the motor neurons. This causes the messages from the brain to stop reaching muscles which leads to weakness and difficulties in doing the normal activities. Musculoskeletal diseases (MSDs) or Myopathy are as bad as motor neuron diseases, as it causes weakness of the muscles. The correctness of disease diagnosis, rehabilitation, and interactional control depend on the high accuracy of the model employed to track muscle activity disorder. In this paper, a new model is introduced for muscle activity disorder tracking through analyzing of Electromyogram signals (EMG). The proposed model integrates wavelet packet transform with candidate selection of Motor unit action potential (MUAP) for efficient feature extraction. A multi-class disorder discrimination of EMG signals is designed for various disorders identification such as Myopathy and Amyotrophic Lateral Sclerosis (ALS). Experiments have been conducted on large dataset of EMG signals to demonstrate the performance and robustness of the proposed model in terms of sensitivity, accuracy, and specificity. The proposed model outperforms existing models. Experimental results indicate up to 2% improvement in terms of Specificity, up to 7.53% improvement in terms of sensitivity in case of binary classification, and up to 5.76 % improvement in terms of accuracy in case of multi-classification.

Through the tremendous increase of users on the microblogging social networks with their associated streams of content, the scarcity of one user’s attention arises. The process of filtering such massive content and discovering who other users could be aligned with his own interests would consume much time. Thus, various mechanisms have been investigated to recommend friends by analyzing the posted content, social graph, or user profiles. In this paper, we propose a new approach for microblog friend recommendation based on the opinion, or sentiment, towards the topics in the microblogs combined with the social graph, in addition to the demographic data available in the user profiles, including age, gender, and location. We have deployed a cloud-based recommender service using R language for big data analytics, which applies our proposed approach to gather feedback from real Twitter users. Results show 0.77 average precision value, with 21 % increase rate considering opinion mining.

Watermarking is one of the most known techniques for authentication, tampering detection, privacy control, etc. Concerning privacy protection of patients' medical records, efforts have been devoted to guarantee the confidentiality of data and medical images during storage and transmission via an untrustworthy channel. Our developed watermarking system, aims at fulfilling such demands through the utilization of a biometric authentication code sender physician's iris code, encrypted patient data and a fuzzy-based Region-of-Interest ROI segmentation algorithm. In this paper, a new hybrid DCT fuzzy biometric-based watermarking scheme has been introduced for privacy protection and source verification of medical and non-medical images. The proposed scheme integrates forward DCT transform with enhanced fuzzy-based ROI regions segmentation. To comply with the imperceptibility and high image quality requirements, the coefficients selection decision depends on perceptual visibility threshold estimation, based upon characteristics of the human visual system HVS. The inclusion of a just-noticeable distortion JND profile, computed for DCT coefficients, is proved to outperform the traditional DCT model, with the major contributions of a new formula for luminance adaptation adjustment and the incorporation of block classification for contrast masking. A block classification has been utilized to differentiate edge regions and thus effectively avoid over-estimation of JND in the selected regions. Moreover, several enhanced fuzzy-based clustering models have been suggested for extraction of ROI regions in medical patterns, which aim at increasing the robustness to noise and yield more homogeneous regions with less spurious blobs. The proposed hybrid fuzzy-based ROI extraction scheme integrates the effect of the local neighborhood and allow it to influence the membership value of each pixel. A new Hybrid FCM H-FCM algorithm is proposed, which integrates spatial information with a 2D adaptive noise removal SS-FCM model. Experiments have been conducted to verify the proposed model. Several attacks have been applied to the proposed scheme and the experiments revealed promising results in terms of visual quality and extracted watermark distortion.

An efficient steganographic method is proposed for embedding secret messages into gray scale and RGB colored images. In this paper, a wavelet domain steganography is adopted for hiding a large amount of data with high security, good invisibility and no loss of secret message. The information is embedded in those areas of the host image that contains high texture to reduce visibility of the embedded information in the host image. The qualified significant wavelet coefficients and their texture and sensitivity to gray value variations across different coarse scales (level 1, level 2 and level 3 wavelet decompositions) are utilized to determine the positions and the magnitudes to adaptively embed the secrete message. We consider the highly textured areas near the edges in high frequency subbands in the wavelet domain. The transformed cover image is divided into number of joint non-overlapping 4 × 4 blocks in each layer (R, G, and B layers). A difference value is calculated from the values of the adjacent pixels in each block, producing the partitioned difference image (PDI). The wavelet subbands are selected to calculate the changes. The selection of the range intervals is based on the characteristics of human vision's sensitivity to gray value variations from smoothness to contrast. The number of bits which can be embedded in each block varies and is decided by the width of the range to which the difference value of the two pixels and block histogram peak (Pk) belongs to. The embedded secret message can be extracted from the resulting stego-image and the original cover image is reversed. The proposed model will produce a high-capacity image steganography technique with acceptable level of imperceptibility and distortion in the cover image and high level of overall security. This solution is independent of the nature of the data to be hidden and produces a stego image with minimum degradation. Experimental results show the feasibility of the proposed methods. Various statistics attacks were also conducted to collect related data to show the security of the method. The experimental results show that the algorithm has a high embedding capacity and a good invisibility. Moreover PSNR of cover image with stego-image shows better results compared with other existing steganography approaches.

In this paper, a new approach has been introduced that integrates an evolutionary-based mechanism with a distributed query sensor cover algorithm for optimal query execution in self-organized wireless sensor networks (WSN). An algorithm based on an evolutionary technique is proposed, with problem-specific genetic operators to improve computing efficiency. Redundancy within a sensor network can be exploited to reduce the communication cost incurred in execution of spatial queries. Any reduction in communication cost would result in an efficient use of battery energy, which is very limited in sensors. Our objective is to self-organize the network, in response to a query, into a topology that involves an optimal subset of sensors that is sufficient to process the query subject to connectivity, coverage, energy consumption, cover size and communication overhead constraints. Query processing must incorporate energy awareness into the system by reducing the total energy consumption and hence increasing the lifetime of the sensor cover, which is beneficial for large long running queries. Experiments have been carried out on networks with different sensors Transmission radius, different query sizes, and different network configurations. Through extensive simulations, we have shown that our designed technique result in substantial energy savings in a sensor network. Compared with other techniques, the results demonstrated a significant improvement of the proposed technique in terms of energy-efficient query cover with lower communication cost and lower size.