ABSTRACT. In many applications related to the acoustic environment, there is a strong need to model or estimate an unknown impulse response, which may be the result of the acoustic coupling between the loudspeaker and microphone of a hands-free communication device. Nevertheless, there are multiple challenges related to such scenarios, since the acoustic impulse responses that need to be identified are time-variant systems, which depend on many factors, e.g., temperature, pressure, humidity, and movement of objects or bodies in the room. In addition, they could have excessive lengths in time (up to hundreds of milliseconds) due to the slow speed of sound in the air, together with multiple reflections caused by the room environment. In this context, a critical issue is related to the identification of long-length room acoustic impulse responses, which are usually modeled as finite-impulse-response filters with hundreds or even thousands of coefficients. In practice, this challenging aspect raises significant difficulties in terms of the complexity and accuracy of the solution. Consequently, it is of major importance to improve the overall performance of the methods and techniques used for the identification of such acoustic systems. In this framework, exploiting the intrinsic nature of the room acoustic impulse responses represent a natural path to follow, which could potentially lead to improved results and further open new perspectives on the identification problem. Recently, we introduced a novel decomposition-based approach for dealing with high-dimension system identification problems. The basic idea is to exploit the Kronecker product decomposition of the long-length impulse responses, together with low-rank approximations, thus solving the system identification problem based on a combination of low-dimension solutions (i.e., shorter filters). The gain of such an approach is important, in terms of both improved performance and reduced complexity. In this keynote, we further present an improved decomposition-based approach for the identification of long-length room acoustic impulse responses, which is a key issue in different challenging scenarios, like acoustic echo cancellation. The main idea is to exploit the intrinsic characteristics of the acoustic impulse response, taking into account its main components, i.e., early reflections and late reverberation, which are very different in nature. Thus, by separately processing these components via the Kronecker product decompositions, their individual low-rank features can be better revealed and exploited. Extensive simulation results obtained in the context of acoustic echo cancellation support the theoretical findings and the performance features of these decomposition-based solutions.

ABSTRACT. Computers and digital platforms are widespread which makes their power consumption very important. Since most dc/dc power stages for computer or telecom applications are currently optimized as modules by manufacturers, the effort in development moves towards controller implementation. While conventional feedback control design methods in frequency domain (lead-lag, or PI-D) are well mastered in industry, with numerous digital or power IC implementation options, this presentation proposes and demonstrates a major leap forward with the use of State Space based design. This time-domain modern control is usually perceived as an advanced control or research topic due to its inherent mathematical support. The advanced concept can however be scaled down for making it easier for either analog or digital implementation by hands-on engineers. Comparatively to PI(D) or lead-lag approaches, the State Space based design guarantees controllability, and provides an easier optimization within a tool like MATLAB®. It is demonstrated to be physically equivalent with cascaded control set around either peak or average current control. Instead of the conventional loop tuning procedures, the simple Ackerman formula calculates gains instantly. Another unique advantage is also proven for the multi-phase dc/dc converters with phase dropping. An in-depth presentation of the method’s actual simplicity, rich in worked examples for either analog or digital implementation, opens this topic to any introductory to intermediate audience.

ABSTRACT. Autonomous vehicles use many technologies and methods to detect and act on surrounding objects. The most common among these technologies is an algorithm called YOLO (You Only Look Once). This algorithm quickly detects objects in an image and classifies these objects accurately. This study examines the use of the YOLO algorithm for signage detection in autonomous vehicles and how this algorithm can be improved. First of all, the basic principles and working mechanisms of the YOLO algorithm are explained. Then, it is explained in detail how this algorithm can be used for plate detection in autonomous vehicles. Various models were trained using the YOLO algorithm and the data set created with real data, and the trained models were tested on real-time systems. Finally, suggestions for the improvement of the YOLO algorithm are presented and how this algorithm can be improved further in the future is discussed.

ABSTRACT. Implementing machine learning algorithms like reinforcement learning in robotics is a continuously changing topic due to continuous tool changes and updates to keep track of new algorithms and tools. This paper presents a new toolchain to implement such algorithms with open-source packages like ROS2 for industrial robots. We developed new tools and procedures to enable the implementation of reinforcement learning algorithms through simulation or controlling real robots with Matlab.

ABSTRACT. This paper presents the collision-free path planning approach for automated guided vehicle (AGV) in an intelligent warehouse environment, optimized by means of recent well-known meta-heuristic algorithms. This novel approach is an assessment and possibilities for path planning and scheduling of the multi-AGVs to complete the given tasks in a minimal travel distance with optimal operation time. Six different metaheuristic algorithms such as PSO, MELGWO, GTO, SFS, MPA and chaotic-improved MPA are compared for the AGVs path optimization capability. In order to test the robustness of the proposed approaches, four different scenarios are presented which include a general obstacle avoidance and three tasks in simple maps that treated as an intelligent warehouse environment. In each scenario, the obstacles are placed in such a way to increase the overall path complexity for an AGV to reach the target destination. The exploration and exploitation phases in the MPA algorithm are enhanced simultaneously by replacing the conventional Gaussian random with chaotic operators to ensure its effectiveness in AGVs path planning optimization. The chaotic MPA algorithm outperforms other algorithms based on the statistical analysis results with overall improvement of 11.0171% in comparison to unoptimized probabilistic roadmap method (PRM) planner. In conclusion, the chaotic MPA algorithm can be efficiently optimized the AGVs path planning in all aforementioned environments.

ABSTRACT. The development of the Industry 4.0 concept has led to a significant gap between the equipment already existing in industrial plants and current technological standards. This gap can adversely affect efficiency and productivity, requiring significant investment in upgrading equipment to keep pace with new technologies. This paper proposes a solution to reduce this gap and at the same time improve the performance of old equipment by integrating a web server into the PLC structure and using I/O-Link devices. This solution aims to transform old equipment into intelligent devices capable of adapting to environmental and production changes without requiring a complete system reconfiguration, which may be a more cost-effective and feasible option than complete replacement. Through the integration of smart features, the upgrade of older equipment to comply with Industry 4.0 standards allows for adaptive functionality without the need for extensive reconfiguration. This transformation facilitates smooth data exchange, real-time monitoring, and augments operational efficiency, productivity, and flexibility.

ABSTRACT. The increase in the share of receivers with non-linear characteristics, among the most common being static frequency converters, produce important voltage and current distortions in electrical networks and require the adoption of effective measures to limit them. The paper presents details regarding the realization and results of an experimental study that reveals how the frequency converters that feed the asynchronous motors influence the electric power quality transmitted to the receivers. The data were acquired and processed in real time with the help of schemes made in the Labview software, being presented in the form of diagrams. The measurements performed in stationary mode highlight the shape and parameters of the line and phase voltages at the output of the converter for various operating frequencies and various load torques, analyze the frequency spectrum and the variation of the coefficients kd and THD. The frequency spectrum of the phase voltage highlights the fact that important harmonics appear around the switching frequency fc, respectively 2fc. Some aspects regarding the importance of using the current filters in improving system performance are also indicated.

ABSTRACT. This paper aims to develop a fast method that can estimate and extract the scenes with crowds from untrimmed streaming videos for surveillance purpose. To this end, we have proposed a simple but effective methodology using regression-based crowd counting algorithms and a custom scene change detector starting from PySceneDetect library. The proposed approach allows to extract crowd scenes from a large amount of data which are useful for crowd statistics, crowd dynamic analysis and crowd scenarios variations detection. The resulted data collection, named UrbanEvent, aims to enhance the performance of semantic segmentation algorithms on crowd scenes. In particular we conduct extensive experiments on crowd counting methods to use as a primary filter on the candidate scenes. The second filter comprises in the determination of a content value threshold, considering the frame-to-frame variety of the candidate changed scene for crowd surveillance purpose. In addition, the experiments show that the K-means clustering method based on feature embedding from ResNet-152 has good results in high and low crowded scenes, being insensitive on context variety, when the proposed method is sensitive to crowd density variations and also to extrinsic camera parameters changes, minimizing the true-negative collected samples, like the car crowded scenes and cars position changes.

ABSTRACT. The purpose of this paper is to analyze Voice over Internet Protocol (VoIP), the wireless network, and the security protocols used in this network environment. VoIP (Voice over Internet Protocol) is a technology that makes it possible to make calls using the Internet connection instead of the analog telephone line. The Voice over Wireless service was developed in early 2003 and was used in locations with many employees, such as hospitals, factories, and office buildings, where employees contact each other through Voice over Wireless LAN services that do not involve an additional management cost.

ABSTRACT. Soccer video analysis is a challenging area of research in computer vision. Several soccer video analysis systems exist for tasks such as player detection and tracking, player performance analysis or team behaviour analysis. These systems are composed of several building blocks, such as image classifiers for shot classification, object detectors for player and ball detection, or object trackers for player and ball tracking. Another useful building block is an image semantic segmentation module, which may be used to segment different elements in the frame. In our setting, it is used to segment the different areas of the soccer field. This paper describes a semantic segmentation network, that segments the 10 different areas of the soccer pitch. This work builds on our previous work for soccer field line segmentation. Our current method directly segments the areas in the image, using a Deep Learning Convolutional Neural Network, based on U-Net, with hierarchical outputs, and balanced or unbalanced loss weights. The hierarchical output contains four outputs with segmentation masks for different segmentation tasks, arranged in a hierarchical tree. Balanced or unbalanced loss weights allow the system's training to be more influenced or less influenced by the accuracy of a particular output. Our best model produces visually convincing results, and is able to achieve an Average Precision of 80.4%, Average Recall of 73.1%, Average F-score of 76.1%, Average Accuracy of 95.1%, and Average IoU of 62.4%.

ABSTRACT. Decomposition-based algorithms have gained much attention lately, in the context of low-rank system identification problems. These algorithms exploit the nearest Kronecker product (NKP) decomposition of the impulse response (usually of long length) and take advantage of low rank approximations. Among them, the recursive least-squares (RLS) algorithm developed in this framework, namely RLS-NKP, has been found to be very suitable in challenging system identification problems that involve long length impulse responses, e.g., like in acoustic echo cancellation. The performance of the RLS-NKP algorithm depends on its decomposition parameter, which is related to the accuracy of low rank approximation. The current paper focuses on the investigation of this aspect and proposes a simple solution for choosing the decomposition parameter, using a preprocessing stage that relies on a low-complexity algorithm. Experiments are performed in the framework of acoustic echo cancellation and the obtained results support the validity of the proposed solution.

ABSTRACT. We propose an analytical design procedure for a Gaussian-shaped circular band-pass 2D FIR filter and also a novel, efficient implementation of the designed filter based on a polyphase decomposition approach. This 2D filter is designed in the frequency domain using a specific frequency mapping applied to a band-pass 1D prototype having a Gaussian shape and imposed specifications (peak frequency, bandwidth). Since we start from a 1D prototype with a factored transfer function, the frequency response of the designed 2D filter also results directly in a factored form. This is a major advantage in actual implementation, since in this case the filter can be synthesized in a cascade structure, and the overall filtering can be achieved sequentially. The proposed implementation is described in a simpler case, namely for a smaller size of the filter kernel and also of the input image. Using the polyphase and block filtering approach, a convenient implementation is obtained for the 2D FIR filter at system level, with a relatively low computational complexity, i.e. a minimum number of arithmetic operations.

ABSTRACT. Pixel-based pixel value ordering (PPVO) represents an efficient solution for low bit-rate and high-fidelity reversible data hiding in digital images. The main novelty of this approach is the selection between two distinct prediction values for each potential host pixel. This paper maintains and refines this feature and adapts it for audio reversible data hiding. For each audio sample value, three possible prediction values are considered: a simple average of neighboring samples around the current position and two weighted averages. Differences between neighboring samples are used to select between the two weighted predictors from the three possibilities. The PPVO framework is then used to select the final predictor for each individual sample from the two remaining predictors. The data embedding algorithm is also adapted to allow for multi-bit embedding, which better exploits the prediction error histogram of audio sample values. Experimental are provided, the proposed PPVO-based approach outperforms other existing audio reversible data hiding scheme, most notably at medium bit-rates.

ABSTRACT. The scientific paper represents original research and analysis of the impact of the use of computerized data deduplication technologies in cases of restoring the Informix databases. Along with the economic growth, a permanent increase in the volume of computerized data was noted, thus showing the need to implement high-performance backup systems in different storage environments. In this context, it was concluded that it is necessary to ensure an improvement of the systems and applications management through which the processes of restoring these volumes of data are ensured. The presented research is a personal point of view of the authors, being argued on the basis of studies and tests carried out in the available computer systems with the aim of streamlining restoration processes and also testing the availability of permanent computer data.

ABSTRACT. Through this paper, we propose an analysis of methods for improving the management of computer data in the context of the need to save, protect and secure them. The implementation of a data saving and restoration system that corresponds to the ISO27001:2018 standard as proposed in this document has the advantage of ensuring an environment in which, in case of disasters, saved data and important applications can be recovered ensuring thus the possibility of continuing the institution's operations in disaster cases. We will present a point of view on the principles and technologies that must be implemented to ensure the restoration of data and the continuation of the operations of intellectual property (IP) organizations after a possible disaster scenario.

ABSTRACT. Currently, a significant category of problems that arise during the evaluation and monitoring process of activities carried out in/for the public interest are related to differences in the understanding of monitoring and evaluation terms. Monitoring is the process of continuously collecting relevant information on how a policy is implemented, while evaluation is the process of using the information obtained during monitoring. In most situations, monitoring and evaluation are seen as simple, periodic tasks, limited in time to simple implementation monitoring reports (such as once a month or every three months, or at the beginning and at the end...) and therefore do not generate essential results and knowledge gained. This paper includes the analysis and presentation of an implementation concept of an IT integration solution between suppliers that also creates a correlation of specific quality indicators for public utility activities.

ABSTRACT. The purpose of application security is to protect software application code and data from malicious action. Application security can be addressed at any stage of the operational process, including the design, development and deployment stages of the IT application, so the scope of what protects application security is constantly growing to adapt to new attack vectors and threat methods. The article aims to perform an analysis of the potential dangers that can affect IT applications to enable management to properly rank the importance of different aspects of information security. Using security access management (SMA) as a tool to measure compliance and risk on the application portfolio in terms of potential exploits, vulnerabilities, misconfigurations and exploitable interfaces is one way to improve cyber security. Another way to improve cybersecurity is the use of multi-factor authentication. SMA can help mitigate risks that are related to insufficient or improperly implemented application security scans, security issues that continue to exist beyond established deadlines for established security standards, and over-privileged or unused access in security tools.

ABSTRACT. The increasing number of photovoltaic systems integrated into the low voltage electric distribution network has raised the concerns of the Distribution Network Operators in the last period. Due to the various technical issues that affect the electric distribution networks, they should pay more attention to the data collected by the smart meters installed to the end-users (consumers and prosumers). This data should process with efficient data-driven algorithms and used in steady-state calculations. This paper proposes a two-stage framework that allows the Distribution Network Operators to analyze the steady state of the low-voltage electric distribution networks incorporating the PV prosumers. It will also help them improve the efficiency of their monitoring and analysis tools. The framework has been tested in a real LV EDN and obtained results confirmed their efficiency.

ABSTRACT. The paper investigates the power losses for LV distribution networks in harmonic polluted state. Actually, the low voltage (LV) distribution networks supply a significant number of nonlinear consumers which causes a distorted and unbalanced state. A negative effect of these states is represented by the increase of power losses in distribution network elements. Even if the symmetrical sinusoidal state is initially considered for the power losses, in reality this often leads to power lines overloading, especially for the neutral. The method for power losses estimation in LV distribution network in harmonic polluted environment is very difficult to use. The aim of this paper is to present a reasonably accurate approach to evaluate the effects of harmonic currents on power losses for LV distribution networks using a clustering-based methodology. Moreover, to highlight the proposed Ward's Hierarchical Clustering Method, a study case to classify the additional power losses due to the distorted state of the LV grid is provided. The proposed approach can be considered as an Expert System to evaluate the influence of harmonic losses in the total LV power losses.

ABSTRACT. Breast cancer is the second most lethal kind of cancer worldwide. On the other side, thermography is a promising technology for early detection of breast cancer. We propose a new model of breast cancer detection in thermal infrared imaging using an ensemble of two Convolutional Neural Networks (CNNs). The considered classifier applies decision fusion according to Dempster-Shaffer theory. The considered CNN modules have an identical architecture, but they use an asymmetric training procedure. The ratio between the number of cancer labeled training thermograms and the normal training thermograms corresponding to first CNN module is denoted by k. The corresponding ratio for the second CNN module is chosen to be (1/k). We have evaluated the influence of the asymmetry training parameter k over the decision fusion classifier performances. One obtains the maximum overall accuracy (OA) of 98.02%, by choosing the parameter k of 1.2.

ABSTRACT. This paper presents an approach of COVID-19 detection for chest Computed Tomography (CT) scan images using an ensemble of two twin Convolutional Neural Networks (CNNs) with the same architecture and training technique. For our investigation model, both CNN modules are intentionally trained with asymmetric volumes of training images belonging to the two considered categories (COVID-19 and NORMAL). Each of the CNN modules has been trained using N chest CT scan COVID-19 labeled images and (N/β) NORMAL labeled images, where β is the class asymmetry parameter defined as ratio of training volumes per class. The classification model uses a simple decision fusion technique, based on maximization of the NET functions in the output layer of each of the two twin CNNs. We have evaluated the influence of the asymmetry parameter β on decision performances. By selecting a small training set of only 142 chest CT COVID images, one obtains best detection performances for β =1.5, corresponding to indices OA of 99.67%, MAR of 0.65% and FAR of 0%.

ABSTRACT. We propose a novel hybrid classifier for hyperspectral images using Bat Algorithm (BA) to optimize the architecture of a Convolutional Neural Network (CNN). BA is applied by minimization of the CNN cross-entropy on the validation set. The objective function is represented as a matrix containing the parameters which define the CNN architecture. The proposed BAT-CNN classifier is evaluated on three hyperspectral datasets: Indian Pines, Pavia University and Salinas. The experiments lead to a better accuracy for the proposed hybrid classifier by comparison to the standalone CNN classifier, for each of the three considered datasets.

ABSTRACT. Emotion classification and sentiment analysis represent crucial research areas within the field of Natural Language Processing. Previous studies have primarily focused on conducting sentiment classification and emotion classification as separate tasks. Only a limited number of researchers have delved into exploring the relationship between these two aspects and invested efforts in deriving one from the other. This study aims to determine sentiment by employing emotion classifications. Specifically, we utilise the ERNIE Tiny deep learning model to classify emotions in Chinese texts, while sentiments are detected through our devised rules. For instance, if emotions such as ‘happiness’ or ‘like’ are present, the sentiment is classified as positive. Conversely, emotions like ‘sadness’, ‘disgust’, ‘anger’, or ‘fear’ classify the sentiment as negative. The experimental results demonstrate the F1 score of 93.00% and 90.14% for positive and negative sentiment, respectively, in Chinese song reviews. These findings substantiate the validity and feasibility of utilising emotions to extract sentiment.

ABSTRACT. This paper presents the results of some experimental studies on the electrical activity of the brain under the conditions of mental arithmetical tasks. It is known that the recognition of evoked potentials from brain bioelectrical signals is a difficult problem mainly due to the cumulative effects of brain waves and the very small level of the signal. The discrimination of states of mental activity specific to arithmetic calculation tasks in relation to the state specific to their absence is investigated by frequency analysis techniques of EEG signals and their classification by topological and temporal criteria. An EEG Biosemi system with 64 electrodes and three basic scenarios for mental tasks that were applied several times to the same subject were used for data acquisition. Each EEG signal was filtered in the dedicated bands δ, θ, α, β, and ϒ. The results in this paper are based on the spectral power analysis for two specific bands: α, β. The evolution of the spectral peaks over time for all 64 recording signals provides a spatial temporal mapping of the electrical activity of the brain to identify the most relevant response channels and identify the most suitable conditions for discriminating mental states using a larger number of electrodes compared to other similar research reported.

ABSTRACT. The conventional loss of life calculation of power transformers is based on IEC and IEEE thermal models that are used already for a long time in the energy sector to evaluate the ageing of transformers and they are integrated in most of the condition monitoring systems. Due to some recent research studies, there was determined a new method to calculate the consumed and remaining lifetime of power transformers, based on a winding insulation resistance model. In this paper, will be presented a study were this new ageing evaluation model will be used to calculate the loss of life of several power transformers installed in the Romanian transmission power grid. The results of this case study will be statistically compared with the results of the conventional thermal model, using data from the condition monitoring systems installed on the power transformer.

ABSTRACT. Prior to the recent work on a low-cost Ateneo mechanical ventilator machine named A-vent, this study demonstrated a simple Digital Twin approach for a real-time monitoring system that can be useful to any mechanical ventilator unit. Previous research concentrated on A-vent design, Near Cloud data caching, and Machine Learning model development. However, it lacks Internet of Things capabilities for remote monitoring applications. This work incorporates new software components to a Near Cloud server that stores and monitors the ventilator and patient data across the wireless network. Wireless sensor nodes attached to the A-vent and patient interaction model capture the time-series waveform of the ventilator, its predictive analysis, and oximeter values. The data queries command displays the data stored in the Near Cloud databases on the monitoring dashboard. It shows a digital representation of the system, allowing real-time updates to be viewed remotely and easily comprehended.

ABSTRACT. Blockchain is a technology that is currently widespread in many fields, it is used to transmit information between two or more parties in a secure way, without the possibility of data loss or interception by an unwanted third party. A blockchain network represents the creation of innovative solutions for the healthcare system, as patient data can be exchanged across hospitals, diagnostic labs, pharmacy firms and doctors. In this paper the topic of blockchain technology and its significant benefits in the field of medical diagnostics is addressed. With the proposed solution a misdiagnosis can be avoided, as there is the possibility to ask for other opinions from other health professionals. The highest level of security information storage allows authorised doctors to access your data from any geographical location, with no territorial limitations. Provides fast, versatile interconnection and enhanced security of patient information. In conclusion, this technology can be incorporated into the current medical system.

ABSTRACT. This paper focuses on the equipment used for fire extinguishing, detection and signaling. The work begins by defining and explaining the importance of these types of equipment in preventing and reducing damage caused by fires. The main types of technologies used in fire extinguishing and detection equipment are presented, highlighting their advantages and disadvantages. Examples of fire-fighting and detection equipment available on the market are also provided, together with tables to compare their characteristics and functionalities. The paper also presents current and future technologies used in fire safety equipment and their potential applications. Finally, the relevant international and local regulations and suggestions for future research directions are provided. In general, the paper aims to provide a comprehensive overview of the latest advances in fire safety equipment and its importance in protecting lives and property.

ABSTRACT. With the time passing the advances in electronics are seriously challenging the natural world in efficiency and organization. Already there are some advanced artificial intelligence achievements that make some scientists asks if it is for the safe of the humanity to be met in practice. Although is seems that are unexplored human brain capacities actually unexplored by common peoples, which would make them superior for a long time in the competition with the electronic brain. The brain functioning relay on some common circuit type. The electronic circuit were not designed in order to copy the neuronal circuits but unavoidable resembling them, as time as are respecting the same natural laws of logical and efficiency.

ABSTRACT. In this study, a low-complexity gait monitoring system using an ESP32 microcontroller, and an MPU-9250 module with an accelerometer, gyroscope, and magnetometer is described. Its performance for binary gait classification using a multilayer perceptron is shown.

ABSTRACT. This paper examines the evolution of encryption techniques, focusing on the first encryption techniques, encryption norms in embedded systems and current encryption techniques. It also addresses encryption norms that are applied in built-in systems, such as AES (Advanced Encryption Standard) and RSA (Rivest-Shamir-Adleman) standards, highlighting their importance in ensuring data security. In addition to presenting existing technologies, the paper also examines a possible direction of encryption techniques in the future. Possible developmental directions such as quantum encryption and machine learning encryption are discussed, providing insight into the future of this area. Finally, the paper synthesizes the information presented, emphasizing the importance of further research and development in the field of encryption, given the constant growth of cyber threats and the importance of protecting information in today's digital world.

ABSTRACT. Due to its desirable characteristics, a generalised multicarrier modulation called Generalized Frequency Division Multiplexing (GFDM), has widespread use in contemporary communication systems. Several methods, such as spectral precoding and pulse shaping, have been developed to shape multicarrier signals' power spectral density (PSD) due to its comparatively high out-of-band radiation. We arrive at a general PSD formulation that is applicable to pulses employed in various subblocks. The wide-sense cyclostationarity and properness statistical hypotheses are the only ones that apply to the vector-valued sequence modulating the active subcarriers. It is discussed whether the based on competence to certain well-known multicarrier systems. The applicability of the GFDM system result is examined.

ABSTRACT. Operation of water distribution systems (WDSs) to reduction of water leakage can be achieved by properly operating variable speed pumps (VSPs). This practical problem can be casted into a nonlinear program (NLP) where decision variables are relative speeds of pumps and or pressure settings of pressure reducing valve while the state variables are flows and nodal heads. The formulated NLP consists of both convex and non-convex constraints, for this reasons, we proposed to apply the sequential convex programming technique (SCP) for solving the NLP. The advantage of the SCP method lies in the fact that instead of solving the non-convex NLP, we solve sequential convex NLPs and the convergent rate of the approach is linear. The effectiveness of the approach was demonstrated on two WDSs for optimal VSP scheduling problem.

ABSTRACT. Water loss reduction in water distribution systems (WDSs) is a challenging task for water utilities worldwide. One of the most reliable and cost- effective ways to reduce water loss is to properly regulate operational pressure of the system through optimizing pressure reducing valve (PRV) placements and operations. This well-known engineering problem can be casted as a mixed-integer nonlinear program (MINLP) where binary variables are introduced to represent positions of PRVs. The existing MINLP model having been used until now for optimizing PRV locations is proven to be hard for MINLP algorithms based gradient methods and therefore is only suitable for small-scale WDSs with a limited number of demand patterns. In this paper, newly non-smooth constraints are proposed to model whether PRVs are placed on links for the optimal PRV locating problem. Since these constraints can be approximated by relaxed forms of complementarity inequalities, high quality MINLP solution can be achieved in a reasonable computation time. To demonstrate the efficiency of our new modeling approach, we compare the performance of the newly formulated MINLP with the existing one for solving optimal PRV locating problems for two case studies. The results have shown that using our new MINLP model, better and more reliable MINLP solution can be found even for WDSs with multiple demand patterns

ABSTRACT. This paper analyzes the vector selection in the predictive current control scheme of a Voltage Source Inverter (VSI) with a Resistor – Inductive load (RL load). Three vector selection ways are considered based on information about the increase and decrease of the load current in the next predictive step. Afterward, the selected vectors will be used in the predictive model. The original predictive current control scheme uses seven vectors of the VSI for the calculation of the predictive model. The vector selection reduces the number of vectors used in the predictive model and also reduces the computational burden of the cost function optimization problem. All three ways are simulated with Matlab/Simulink and compared with the original predictive current control scheme. This selection can minimize computational burden, but the quality of the load currents can become poor. However, suppose the number of vectors in the vector selection way is three. In that case, the quality of the load currents of this method is equivalent to the quality of the load currents of the original predictive current control scheme.

ABSTRACT. Histology analysis is currently a gold standard in analyzing cancer. Nuclei segmentation is vital in histopathology analysis. However, it is challenging due to limited data and extreme conditions. With the advent of transfer learning methods, the solution to this problem is possible. We propose a transfer learning-based approach for segmenting the nuclei in Head and Neck (H&N) cancer histology images. The suggested technique comprises two stages. In the first stage, we train our previously proposed architecture, DAN-Nuc Net, on generic histology data to achieve generic nuclei segmentation in histology. We use the PanNuke dataset, which has over 8000 histology images, to train DAN-Nuc Net. In the second stage, we use transfer learning techniques to optimize our network for two types of histology mages, i.e., Hematoxylin and Eosin (H&E), and P63 independently. Selected deep layers of the pre-trained DAN-Nuc Net are frozen. Then the model is re-trained on the new datasets. Compared to the state-of-the-art, our method has shown superior performance in DSC and JI (0.8702 and 0.7596).

ABSTRACT. Lithium-ion batteries (LBs) have become increasingly popular for use in electric vehicles, aircraft, and portable electronic devices due to their high-energy storage capacity and extended lifespan. As a result, the demand for Li-ion batteries has risen significantly compared to other rechargeable batteries. During normal working conditions, any fault in the battery may lead to severe damage to equipment or human. As a preventative measure, developing a Prognostic and Health Management (PHM) system that can detect faults early on is essential. PHM systems can provide early warning of faults and improve reliability and safety. A Prognostic and Health Management (PHM) system for batteries comprises three components: determining the State of Charge (SOC), the State of Health (SOH), and the Remaining Useful Life (RUL). This paper will explore deep learning (DL) techniques to predict the SOC, SOH, and RUL of batteries. Generally, DL based method for PHM has four main stages, data collection, extraction of features, training, and testing. DL-based techniques for PHM of LB will be discussed in detail and also make comparisons to understand the effectiveness.

ABSTRACT. Unmanned aerial vehicles (UAVs) are one of the applications of device-to-device (D2D) technology, where two UAVs can communicate with or without the existence of a base station. It facilitates high content delivery, low latency, low cost, and high data rate communication in wireless networks. Despite the aforementioned benefits, an eavesdropping attack is one of the critical challenges that cause interference in UAV communication in the presence of an eavesdropper. To mitigate the aforementioned challenges, we proposed a secure resource allocation for UAV communication in wireless networks. The applied single carrier orthogonal frequency division multiple access (SC-OFDMA) technique utilize the orthogonality feature to provide secure UAV communication in wireless networks. Moreover, a zero-sum game approach is considered to enable the secure and optimal resource allocation for UAV communication. The main aim of the zero-sum game theory is to maximize the communication channel’s data rate and secrecy capacity to achieve a secure and optimal UAV communication scenario in wireless networks. Finally, the performance and simulation of the proposed approach are evaluated against the existing random approach in terms of various metrics such as data rate and secrecy capacity of the communication channel.

ABSTRACT. The rising data volumes force significant bottlenecks on the 6G for IoT (6G-IoT) network management functions, which limits the control, flexibility, and interoperability among devices, protocols, and end applications. Solutions like software- defined networking (SDN), and network function virtualization (NFV) are proposed with 6G, but the core management operations are still manual. Thus, to automatically upscale these 6G-IoT networks at reduced cost orchestration complexity, zero- touch networks (ZTN) are proposed. ZTN in 6G-IoT allows a high degree of automation and seamless integration of services. The article proposes a scheme, Zero-Load, that integrates ZTN at the core routing functionality of the 6G-IoT applications. We present a load balancing and traffic classification scheme through the ZTN networking stack for core routers. The ZTN router configuration fabric connects applications with the core services. Further, we present a Gaussian kernel-based support vector machine (SVM) classifier at the ZTN automation layer, which classifies the normal traffic and attack traffic. The proposed work is compared for parameters like mean time to response (MTTR), and resolution latency against baseline SDN and NFV schemes. Using ZTN, an average improvement of 32.45% is obtained in MTTR, 87.89% in resolution latency (against a query). Using the Gaussian RBF kernel, an accuracy of 0.9914 is reported. These results indicate that ZTN-based management paves the way toward a more dense and intelligent 6G-IoT network.

ABSTRACT. This paper investigates the impact of feature encoding techniques on the explainability of XAI (Explainable Artificial Intelligence) algorithms. Using a malware classification dataset, we trained an XGBoost model and compared the performance of two feature encoding methods: Label Encoding (LE) and One Hot Encoding (OHE). Our findings reveal a marginal performance loss when using OHE instead of LE. However, the more detailed explanations provided by OHE compensated for this loss. We observed that OHE enables deeper exploration of details in both global and local contexts, facilitating more comprehensive answers. Additionally, we observed that using OHE resulted in smaller explanation files and reduced analysis time for human analysts. These findings emphasize the significance of considering feature encoding techniques in XAI research and suggest potential for further exploration by incorporating additional encoding methods and innovative visualization approaches.

ABSTRACT. Abstract—Image security is a critical issue in the recent era of multimedia applications. Mostly, there has been an exponential growth in image data communication among various parties. Due to exchange of confidential image data via communication channels, the attackers may intercept the communication channel that led to serious security breaches. In this manner, the confidential data in the digital images may be illegally misused by the unauthorized adversaries. To counter these attacks, one of the viable solutions is to apply image cryptography algorithms for secure communication between sender and receiver. The existing image cryptography methods exhibit a trade-off between security and performance. In this paper, we expound a lightweight image cryptography algorithm for the encryption of digital images. Our approach makes use of simple linear bitwise operator along with various encryption keys that is shared between sender and receiver. The proposed approach is efficient in terms of complexity and highly secure to send data among different users. The method exhibits a, mean square error (MSE) of 13218.83, peak signal-to-noise ratio (PSNR) of 7.364 and signal-to-noise ratio (SNR) of 16.864 respectively.

ABSTRACT. Due to the digital era and rapid advancements in technology, an enormous amount of publicly available information is generated. Open Source Intelligence (OSINT) is a term used to describe the search, collection, analysis and use of information from open sources. OSINT has emerged as a valuable tool in cybersecurity, transitioning from its basic usage in social engineering to becoming a powerful tool in red teaming exercises and seamlessly integrating into the fabric of digital forensics. OSINT environment is constantly evolving with the emergence of new use cases and complex tools. This paper presents MTAFinder, an application for automated data collection that integrates OSINT open source tools and free online services. The platform is modular, scalable and provides a web API designed to gather data about people, phone numbers, web domains, e-mail addresses and IP addresses.

ABSTRACT. Iris recognition systems are attaining wide consideration for securing critical computing infrastructure. Though, these systems are exposed to a various kind of attacks or threat that may degrade the performance of the overall system in real-time scenario. Among all, spoofing attacks are the most attempted once, where an adversary tries to break the security by offering a fake artifact of a real biometrical trait. In this paper, we explore most recent development in the area of iris vitality detection mechanisms that are inspired by deep learning-based models. Besides, we illustrate an analysis of iris vitality detection (IVD) anti-spoofing datasets and performance evaluation protocols. Our overall instigation indicates that majority of the recent methods perform well in known attack scenario, however exhibit limited performance in unseen attacks set-up. One of the major challenges is to develop a robust IVD method with smaller datasets that has demonstrates superior performance in cross-scenario settings.