A review of modern methods and technologies for automatic machine translation for the deaf and hard of hearing is presented, including recognition and synthesis of both spoken and sign languages. These methods aim to facilitate effective communication between deaf/hard-of-hearing and hearing individuals. The proposed solutions have potential applications in contemporary human-machine interaction interfaces. Key aspects of new technologies are examined, including methods for sign language recognition and synthesis, audiovisual speech recognition and synthesis, existing corpora for training neural network models, and current systems for automatic machine translation. Current neural network approaches are presented, including the use of deep learning methods such as convolutional and recurrent neural networks as well as transformers. An analysis of existing corpora for training recognition and synthesis systems is provided, along with an evaluation of the challenges and limitations of existing machine translation systems. The main shortcomings and specific problems of current automatic machine translation technologies are identified, and promising solutions are proposed. Special attention is given to the applicability of automatic machine translation systems in real-world scenarios. The need for further research in data collection and annotation, development of new methods and neural network models, and creation of innovative technologies for processing audio and video data to enhance the quality and efficiency of the existing automatic machine translation systems is highlighted.

This paper examines routing algorithms for networks on a chip (NoC). An analysis of existing routing algorithms is provided; their limitations and areas of application are highlighted. The algorithms were evaluated taking into account the requirements of specific applications and architecture features. The results of comparing the performance of the considered algorithms are presented. Analysis and comparison of various routing algorithms for NoC are carried out taking into account critical characteristics. The main attention is paid to such routing algorithms as the deterministic XY algorithm, the model rotation algorithm, congestion-aware routing, fault-tolerant routing, Quality of Service routing, and the ant colony algorithm. It is shown that the choice of routing algorithm should be based on the specific requirements and conditions of use of the network. The importance of adapting to a variety of conditions and tasks that NoC users and developers may encounter is shown. Based on data from existing studies, an analysis of algorithms was carried out based on several key indicators: latency, throughput, adaptability, fault tolerance and implementation complexity. The strengths and weaknesses of each algorithm are identified in various use scenarios and under different network loads. It is shown that the choice of a routing algorithm should be based on the specific requirements and conditions of use of the network, as well as on the balance between performance, adaptability, fault tolerance and implementation complexity. The study contributes to the understanding of the effectiveness of various routing algorithms in NoC, providing recommendations for their selection depending on the specific application requirements and system architecture. The study contributes to a better understanding of the impact of routing algorithms on the overall performance of NoC, suggesting directions for further improvements in this area. The results of the work can be applied in the design and development of highperformance multiprocessor systems on a chip where efficient data routing between various systems components is a key factor in ensuring high performance. The importance of developing fault-tolerant routing algorithms that can ensure the continuity of system operation in the event of failures of individual components or units is emphasized. This is especially important for mission-critical applications where service continuity and reducing the risk of data loss are top priorities.

The paper presents the results of an experimental study of the possibility of using a narrow-band semiconductor distributed feedback laser diode used as a source and detector of optical radiation to detect the spectral response from a fiber Bragg grating. The DFB laser “LDI-1550-DFB-2.5G-20/70” from the company “Laserscom”, mass-produced on the Russian market and having standard characteristics, was chosen as the laser diode under study. To sweep the central wavelength of a semiconductor distributed feedback laser diode in the range 1549.5–1552 nm, direct pulse current modulation was used with a frequency of 100 kHz, a duty cycle of 40, and a current value of 1 A per pulse. The radiation reflected from the fiber Bragg grating corresponding to the central Bragg wavelength was recorded as a change in voltage at the anode and cathode of the laser diode due to the photoelectric effect in the laser diode. An experimental assessment of the optoelectronic parameters of a laser diode in photovoltaic and short-circuit modes was carried out: dark current, bandwidth and spectral sensitivity. The evaluation was carried out at a temperature of 25 °C. A measuring circuit has been created to detect the response from a fiber Bragg grating based on direct pulse current modulation and the photovoltaic mode of a semiconductor distributed feedback laser diode. It is shown that the photovoltaic mode of the laser diode is applicable to problems of recording optical radiation. The amplitude-frequency characteristic of a laser diode in the photovoltaic mode was experimentally obtained depending on the forward bias voltage. It is experimentally found that the –3 dB bandwidth is 300 MHz and the maximum sensitivity is 0.1 A/W in short-circuit mode, and the amplitude response is linear in the wavelength range from 1540 to 1560 nm. For the laser diode under study, the reverse branch of the current-voltage characteristic was experimentally obtained and the dark current at zero bias of the laser diode is 12.5 pA. The demonstrated method of FBG interrogation can be used for miniaturization and simplification of optical devices for fiber Bragg grating interrogation. The obtained results may be useful to specialists in fiber optic sensors, system for interrogation and processing signals from fiber optic sensors.

The results of investigation of the gain properties of 1300 nm vertical-cavity surface-emitting lasers active regions based on In0.60Ga0.40As/In0.53Al0.20Ga0.27As superlattices and threshold characteristics comparison of superlattices and highly lattice mismatched In0,74Al0,16Ga0,10As quantum wells are presented. The heterostructure of injection lasers with an In0.60Ga0.40As/In0.53Al0.20Ga0.27As superlattice was grown by molecular beam epitaxy. Mesa structure of injection lasers was obtained by selective liquid etching followed by the application of ohmic contacts. The formation of injection lasers with various cavity lengths is performed using the method of manually cleaving mirrors. The output characteristics were measured in a pulsed mode using a large area calibrated germanium photodiode. Spectral characteristics were measured using a spectrophotometer based on monochromator. The achieved threshold characteristics (modal gain about 40 cm–1, transparency current density about 650 A/cm2, internal optical losses about 8 cm–1) of injection lasers based on In0.60Ga0.40As/In0.53Al0.20Ga0.27As superlattices with low lattice mismatch InGaAs layers are comparable to previously presented lasers based on active regions with strongly strained In0,74Al0,16Ga0,10As quantum wells. The characteristic temperatures T0 and T1 were 60 K and 87 K for injection lasers with a cavity length of 1 mm. An increase in the frequency of small-signal modulation of vertical-cavity surface-emitting lasers and their temperature stability is associated with the use of highly strained In0.60Ga0.40As/In0.53Al0.20Ga0.27As superlattices. The proposed active regions based on InGaAs-InP superlattices have the potential to be used in the development of vertical-cavity surface-emitting lasers in the 1300 nm spectral range. The findings of this work can be applied in the realization of experimental species and optimization of modulation parameters for vertical-cavity lasers operating in the 1300 nm wavelength range. 

One of the main goals of jewelry is to give the product an aesthetic and artistic look. This can be achieved by changing its color. The promising methods of precious metal coloration is a one-step laser method of forming nanostructures with plasmonic properties. However, the lack of understanding of the mechanism of formation of color coatings remains an unresolved problem today. Surface structures are usually considered as a set of individual spherical nanoparticles. But to fully understand the physicochemical processes taking place, it is necessary to consider nanoparticles in aggregate as agglomerates of these nanoparticles on the surface. The silver samples of 99.99 % purity were selected for the study. Laser exposure was carried out in air using a system based on an ytterbium fiber laser with nanosecond pulse duration. The silver surface was processed by line-by-line scanning along one and two axes with a focused laser beam with the diameter d0 = 50 μm. Optical and scanning electron microscopy were used to characterize the silver surface before and after laser treatment. In this work, the effect of some laser exposure parameters, such as laser pulse energy and pulse repetition rate, on the optical properties of silver surface were investigated. The focus of this work is on generated laser-modified surface nanostructures and the character of their change when going from single axis scanning to line scanning. It is shown that surface topography changes are also observed in the region outside the immediate treatment zone. The uneven distribution of nanostructural elements on the surface of the treated area is registered, which causes the irregularity of the observed surface color at the microlevel. Based on the analysis of the obtained data, a hypothesis of nanostructure formation is proposed. Under laser exposure, individual spherical-shaped nanoparticles are initially formed on the silver surface. Then with increasing temperature their concentration increases significantly. This leads to their adhesion and formation of irregularly shaped clusters of agglomerated nanoparticles. The obtained new data on the process of formation of surface nanostructures allow us to expand the understanding of the ongoing processes, as well as to approach the integration of the method of direct laser coloring of silver in the jewelry industry.

One of the problems solved by developers of autonomously controlled unmanned aerial vehicles is the task of determining by the drone its exact position over the terrain without the help of global satellite navigation systems. The existing mass-dimensional and energy limitations for small-sized drones lead to the necessity of using relatively simple algorithms in drone computing devices. The paper considers methods of navigation of unmanned aerial vehicles using computer vision implemented by on-board optical and computing devices. Machine vision implemented by on-board computing devices provides autonomy of small-sized aircraft in the absence or unstable communication channel with the control center and/or satellite navigation system. The proposed algorithm solves the problem of identifying an area of terrain observed from a drone with a terrain image stored in the memory of the drone control system. The drone location is determined by the minimum (maximum) value of the discrepancy between the observed current image and the image of the terrain area stored in the drone memory device. The solution of the identification problem is based on the concept of immunocomputing using singular value decomposition of the feature matrix of the identified objects. This approach allows providing high quality indicators of identification due to decomposition of the feature matrix into three simple transformations for transition to a new feature space which is not identifiable, but whose components are statistically significant. The quality indicators of the developed algorithm were evaluated in comparison with the known method of image identification by calculating the correlation function between two arrays of features. A series of tests were carried out in which the probability of correct location determination and the speed of the algorithms were evaluated for the same initial data. It is shown that when pre-preparing a “reference” image stored in the drone memory device, the speed of the developed method exceeds the speed of the method based on the calculation of the correlation function of the compared images by an order of magnitude. The mean absolute error of correct positioning using the proposed method ranges from 0.109 to 0.153. The proposed algorithm can be used by developers of navigation systems for small-sized unmanned aerial vehicles due to its low resource requirements while maintaining a level of accuracy sufficient in the context of solving problems of orientation on the terrain. Devices realizing the proposed orientation algorithm have better energy and mass-size characteristics. 

The paper presents the concept of a point amplitude sensor for the registration of displacement of geotextile, a synthetic fabric that is used to reinforce geotechnical structures such as a dam. The implementation of a system for continuous monitoring of the structural condition of a building based on the concept of a “smart” geotextile has the potential to significantly enhance the safety of the structure. Such a system could provide early warning of the necessity for unscheduled repairs, the occurrence of an emergency situation, and the need for the immediate cessation of building operations, evacuation of personnel or population. The capabilities of existing technical solutions for displacement sensors have been evaluated. It is not feasible to apply existing monitoring systems utilizing fiber Bragg Grating Sensors (FBG) in the context of geotextile. This is due to the greater pliability of the soil which exhibits minimal elastic deformation. In addition, FBG sensors are much more expensive in production compared to telecommunication optical fiber. The single-mode fiber which constitutes the sensing element, forms one or more loops that are placed between movable stops that are attached to the sensor body and to the movable activator. At the point of macro bending of the reinforcing fiber, the phenomenon of total internal reflection is disrupted, which in turn gives rise to amplitude modulation of the radiation. The macro bending is proportional to the displacement of the activator attached to the geotextile. This paper presents the design, dimensions and mathematical relationships of the sensing element as well as the dimensions and characteristics of the design elements for signal processing. The sensor model is constructed from ABS plastic and fiber Corning SMF-28. An experimental setup was constructed to test the proposed concept which involved controlling the displacement of the activator, the input and output of radiation. The dependences of the output power on the fiber bending diameter, ranging from 25 to 11 mm, and the displacement, up to 14 mm, at a radiation wavelength of 1550 nm, were determined. It was demonstrated that the obtained dependences were monotonic and exhibited quasi-linear plots. The kinks at the small diameter of the fiber bend are caused by two factors: the intensive radiation output from the core to the cladding and scattering within it; and at the large diameter, they are due to small bending losses. The conducted studies have demonstrated that the sensor is capable of reliably detecting displacements up to 0.5 mm. The results exhibited good repeatability. The proposed sensor demonstrated inferior accuracy compared to FBG sensors. Conversely, at comparable accuracy of ground displacement registration, the proposed sensor was observed to be an order of magnitude more cost-effective than FBG sensors.

A new control algorithm for nonlinear plants is proposed, ensuring the controlled variable stays within a given set under conditions of parametric uncertainties, external disturbances and high-frequency noises in measurements. The problem is solved in two stages. In the first stage, a low-pass filter is applied to eliminate high-frequency components in the measured controlled signal. In the second stage, a coordinate transformation represents the initial problem with given restrictions as an input-state stability analysis problem of a new system without constraints. An output feedback algorithm has been developed for uncertain nonlinear systems under conditions of parametric uncertainties, external disturbances, and high-frequency noise in measurements. Simulations in MATLAB/Simulink are given. The simulation results show the efficiency of the proposed algorithm. The proposed algorithm can effectively solve control problems for power systems or electromechanical systems in the presence of measurement noises.

Polymer brushes, as modifying coatings, significantly improve the tribological properties of various contacting surfaces. The friction that arises when an external load is applied and the polymer brushes laterally shift relative to each other is determined by their interaction energy and the depth of interpenetration. If the brushes are immersed in a lowmolecular-weight solvent, the friction force can be controlled by varying the solvent quality through changes in external conditions, such as temperature, chemical composition of the solution, and so on. It should be noted that theoretical studies on the effect of solvent quality on the tribological properties of brushes are practically absent. To determine the influence of solvent quality on the interaction of flat polymer brushes, two complementary approaches were used: analytical and numerical self-consistent field methods. In both cases, a coarse-grained model of polymer brushes was employed. The solvent quality in the model was defined through the Flory-Huggins parameter for polymer-solvent interaction. A quantitative assessment of the overlap zone width, osmotic pressure, and friction force arising when the brushes approach each other was conducted. A theoretical description of the friction force in the low shear rate regime was proposed based on the Brinkman equation for two compressed brushes sliding against each other. It was shown that, with constant total polymerization degree, grafting density, and lateral sliding speed of the brushes relative to each other, the width of the overlap zone decreases following a power law with increasing inter-plane distance, regardless of solvent quality. Under conditions of strong compression of flat polymer brushes, the friction force approaches a certain limiting value, while the friction coefficient tends to zero, independent of solvent quality. In the moderate pressure region, the friction coefficient significantly increases with a decrease in the solubility of the grafted polymers under the same applied external load and the composition of the flat polymer brushes. The analytical method showed high agreement with the data from the numerical simulations. The obtained results allow for predicting the tribological properties of polymer brushes depending on solvent quality and, consequently, predicting the effect of external conditions on the friction force between modified surfaces.

The task of automatic metainformation recognition from audio sources is to detect and extract data of various natures (speech, noises, acoustic scenes, acoustic events, anomalies) from a given audio input signal. This area is well developed and known to the scientific community and has various approaches with high quality. But, the vast majority of such methods are based on large neural networks with a huge number of weights to be trained. Subsequently, it is impractical to use them in environments with severely limited computing resources. The smart device industry is currently growing rapidly: smartphones, smart watches, voice assistants, TV, smart home. Such products have limitations in both processor and memory. At that moment, the State-of-the-Art way to cope with these conditions is to use so-called low-complexity models. Moreover, in recent years, the interest of the scientific community in the above-mentioned problem has been growing (DCASE Workshop). One of the most crucial subtasks in the global meta information recognition problem is the task of Automatic Scene Classification and the task of Sound Event Detection. The most important scientific questions are the development of both the optimal low-complexity neural network architecture and learning algorithms to obtain a low-resource, high-quality system for classifying acoustic scenes and detecting sound events. In this paper the datasets from DCASE Challenge “Low-Complexity Acoustic Scene Classification” and “Sound Event Detection with Weak Labels and Synthetic Soundscapes” were used. A multitask neural network architecture was proposed consisting of a common encoder and two independent decoders for each of the two tasks. The classical algorithms of multitask learning SoftMTL and HardMTL were considered, and their modifications were developed: CrossMTL, which is based on the idea of reusing data from one task when training the decoder to solve the second task, and FreezeMTL, in which the trained weights of the common encoder are frozen after training on the first task and used to optimize the second decoder. As a result of the experiments, it was shown that the use of the CrossMTL modification can significantly increase the accuracy of the classification of acoustic scenes and event detection in compare with classical approaches SoftMTL and HardMTL. The FreezeMTL algorithm made it possible to obtain a model that provides 42.44 % accuracy in scene classification and 45.86 % accuracy in event detection, which is comparable to the results of the baseline solutions of 2023. In this paper, a low-complexity neural network consisting of 633.5 K trainable parameters was proposed, requiring 43.2 M MACs to process one second audio. This approach uses 7.8 % fewer trainable parameters and 40 % fewer MACs compared to the naive application of two independent models. The developed model can be used in smart devices due to a small number of trainable parameters, as well as a small number of MACs required for its application.

As neural networks become more complex, the number of parameters and required computations increases, which complicates the installation and operation of artificial intelligence systems on edge devices. Structural distillation can significantly reduce the resource intensity of using any neural networks. The paper presents a method for optimizing neural networks that combines the advantages of structural distillation and a genetic algorithm. Unlike evolutionary approaches used to search for the optimal architecture or distillation of neural networks, when forming distillation options, it is proposed to encode not only the parameters of the neural network, but also the connections between neurons. The experimental study was conducted on the VGG16 and ResNet18 models using the CIFAR-10 dataset. It is shown that structural distillation allows optimizing the size of neural networks while maintaining their generalizing ability, and the genetic algorithm is used to effectively search for optimal distillation options for neural networks, taking into account their structural complexity and performance. The obtained results demonstrated the effectiveness of the proposed method in reducing the size and improving the performance of networks with an acceptable loss of quality

The main problem in building intelligent systems is the lack of data for machine learning, which is especially important for sign language recognition for the deaf and hard of hearing. One of the ways to increase the amount of data for training is synthesis. Unlike speech synthesis, it is impossible to create a sequence of gestures in Vietnamese and some other languages that exactly repeat the text. This is due to the significant limitations of the gesture dictionary and the different word order in sentences. The aim of the work is to enrich the educational corpus of video data for use in creating recognition systems for the Vietnamese Sign Language (ViSL). Since it is impossible to translate the words of the source text into gestures one to one, the problem of translating from a regular language into a sign language arises. The paper proposes to use a two-phase process for this. The first phase involves pre-processing the text with standardization of the text format, segmentation of words and sentences, and then encoding the words using the sign language dictionary. At this stage, it should be noted that there is no need to remove punctuation marks and stop words, since they are related to the accuracy of the N-gram model. Next, instead of using syntactic analysis, a statistical method for forming a sequence of gestures is used, and the Markov model on the transition graph between words is taken as a basis in which the probability of the next word depends only on the two previous words. Transition probabilities are calculated on the existing marked corpus of the ViSL. The Breadth-first Search method is used to compile a list of all sentences generated based on a given grammatical rule and a matrix of semantic interactions between words. The inverse of the logarithm of the product of the probabilities of co-occurrence of consecutive 3-word phrases in a sentence is used to estimate the frequency of occurrence of that sentence in a given data set. Based on the ViSL data of 3,234 words, we calculated probability matrices representing the relationships between words based on Vietnamese natural language data with 50 million sentences collected from Vietnamese newspapers and magazines. For different grammar rules, we compare the number of generated sentences and evaluate the accuracy of the 50 most frequent sentences. The average accuracy is 88 %. The accuracy of the generated sentences is estimated by manual statistical methods. The number of generated sentences depends on the number of word parts that are labeled according to the grammar rules. The semantic accuracy of the generated sentences will be very high if the search words are labeled with the correct part-of-speech tagging. Compared with machine learning methods, our proposed method gives very good results for languages without inflections and word order that follow certain rules, such as Vietnamese, and does not require large computational resources. The disadvantage of this method is that its accuracy largely depends on the type of word, sentence, and word segmentation. The relationship of words depends on the observed dataset. Future research direction is to generate paragraphs in sign language. The obtained data can be used in machine learning models for sign language processing tasks.

Detection and handling of anomalous behavior in the network systems are peremptory efforts to ensure security for vulnerable infrastructures amidst the dynamic context of cybersecurity. In this paper, we propose an ensemble machine learning model architecture that leverages the strengths of XGBoost, Gradient Boosting, Random Forest, and Support Vector Machine models to identify anomalies in the dataset. This method utilizes an ensemble of these models with weighted voting based on accuracy to enhance anomaly detection for robust and adaptive real-world network security. The proposed ensemble learning model is evaluated on standard metrics and demonstrates exceptional efficacy, achieving an impressive accuracy of 99.68 % on NSL KDD dataset. This remarkable performance extends the model prowess in discerning anomalies within network traffic showcasing its potential as a robust tool for enhancing cybersecurity measures against evolving threats. 

Attribute Based Access Control (ABAC) is one the most efficient, scalable, and well used access control. It’s based on attributes not on users, but even when the users want to get access to some resource, they must submit their attributes for the verification process which may reveal the privacy of the users. Many research papers suggest blockchain-based ABAC which provides an immutable and transparent access control system. However, the privacy of the system may be compromised depending on the nature of the attributes. A Zero-Knowledge Proof, Ethereum-Based Access Control (ZK‑ABAC) is proposed in this paper to simplify the management of access to the devices/objects and provide an efficient and immutable platform that keeps track of all actions and access management and preserve the privacy of the attributes. Our ZK-ABAC model utilizes smart contracts to facilitate access control management, Zero-Knowledge Succinct NonInteractive Argument of Knowledge (ZK-SNARK) protocol to add privacy to attributes, InterPlanetary File System (IPFS) network to provide distributed storage system, and Chainlink to manage communications and data between on/ off-chain systems. Comprehensive experiments and tests were conducted to evaluate the performance of our model, including the implementation of ZK-SNARK on the Ethereum blockchain. The results demonstrated the scalability challenges in the setup and proving phases, as well as the efficiency gains in the verification phase, particularly when scaled to higher numbers of users. These findings underscore the practical viability of our ZK-ABAC model for secure and privacy-preserving access control in decentralized environments.

The study aimed to improve the efficiency of detecting and mapping felling using satellite imagery, in order to identify violations of environmental regulations. Traditional remote sensing data interpretation methods are labor-intensive and require high operator expertise. To automate the satellite image interpretation process, numerous approaches have been developed, including those leveraging advanced deep machine learning technologies. The presented work conducted a comparative analysis of convolutional and transformer neural network models for the segmentation of felling in summer Sentinel-2 satellite imagery. The convolutional models evaluated included U-Net++, MA-Net, 3D U-Net, and FPN-ConvLSTM, while the transformer models were SegFormer and Swin-UperNet. A key aspect was the adaptation of these models to analyze pairs of multi-temporal, multi-channel satellite images. The data preprocessing, training sample generation, and model training and evaluation procedures using the F1 metric are described. The modeling results were compared to traditional visual interpretation methods using GIS tools. Experiments on the territory of the Khanty-Mansiysk Autonomous Okrug showed that the F1 accuracy of the different models ranged from 0.409 to 0.767, with the SegFormer transformer model achieving the highest performance and detecting felling missed by human interpretation. The processing time for a 100 × 100 km2 image pair was 15 minutes, 16 times faster than manual methods — an important factor for large-scale forest monitoring. The proposed SegFormer-based felling segmentation approach can be used for rapid detection and mapping of illegal logging. Further improvements could involve balancing the training dataset to include more diverse clearing shapes and sizes as well as incorporating partially cloudy images.

The active development of digital technologies, Internet of Things technologies, and virtual tests requires an increase in the volume of information collected and used, which is placed in data storage systems. The rapid growth of data volume leads to stricter requirements for storage. One of the main requirements for storage is to increase the reliability of storing large amounts of information. This implies the need to assess the reliability of storage equipment. For these purposes, it is necessary to evaluate such reliability indicators as the probability of failure-free operation, the probability of failures, the average residual resource, and the gamma percent resource. Traditionally, reliability indicators are evaluated with an exponential distribution of failure time. In a real situation, the samples of failure times of storage equipment are small, for which it is impossible to uniquely identify the initial distribution. In this article, a model is proposed for evaluating reliability indicators as a gamma percent residual resource in conditions of incomplete data presented by small samples of random variables of equipment uptime. The scientific novelty of the presented work consists in obtaining a general solution to the problem of determining the guaranteed gamma percent residual life of equipment in conditions of incomplete data presented by small samples of developments before equipment failure. The mathematical formalization of the problem of estimating the gamma percent residual life of storage equipment in conditions of incomplete data presented by small samples is performed in the form of a stochastic equation model, the solution of which is a guaranteed (lower, upper) estimate of the gamma percent residual life of equipment. A model for estimating the gamma percent residual life of storage equipment in conditions of incomplete data is presented. In the general case, the problem of finding guaranteed (lower and upper) estimates of the gamma percent residual life of equipment on a set of functions for the distribution of uptime of equipment with specified moments equal to sample moments determined from small samples is solved. At two points in the uptime of the equipment, analytical ratios were obtained to determine the gamma percent residual life. The performance of the model is demonstrated by the example of determining the lower guaranteed estimate of the gamma percent residual resource of the HP EVA P6500 disk array model. The results obtained can be used by specialists in evaluating and optimizing the gamma percent residual life of storage equipment. 

Multiple Sclerosis (MS) is a progressive autoimmune disease affecting the central nervous system, causing communication disruptions between the brain and the body. Early and accurate detection of MS lesions in brain Magnetic Resonance Imaging (MRI) scans is crucial for effective treatment. This paper proposes MSNet, a deep learning-based approach for automatic detection and diagnosis of MS lesions from MRI images, leveraging Convolutional Neural Networks (CNNs) for precise lesion identification and classification. Our methodology involves a comprehensive analysis of MRI datasets, including preprocessing steps such as normalization and lesion segmentation. We propose a novel CNN architecture tailored for MS lesion detection, achieving an accuracy rate of 98.2 % on the test dataset. By incorporating advanced image recognition techniques, our system classifies MS lesions from diverse brain pathologies present in MRI images. The model also highlights MS lesions within the MRI images, aiding neuroradiologists in accurate diagnosis and treatment planning. This study contributes significantly to improving MS diagnosis by providing a reliable and automated tool for lesion detection and classification.

The development of digital communication systems is associated with the increasing number of disruptive behavior incidents that require rapid response in order to prevent negative consequences. Due to weak formalization of human aggression, machine learning approaches are the most suitable for this area. Machine learning approaches require representative sets of relevant data for efficient aggression recognition. Datasets developing implies such problems as dataset labels relevance to the real behavior, the consistency of the situations, where behavior is manifested, and the naturalness of behavior. The purpose of this work is the development of an aggressive behavior datasets creation methodology that reflects the key aspects of aggression and provides relevant data. The work reveals the developed methodology for creation of multimodal datasets of natural aggression behavior. The analysis of human aggression subject area substantiates the key aspects of human aggression manifestations (the presence of subject and object of aggression, the destructiveness of the aggressive action), the behavior analysis units — the time intervals of audio and video with the localized informants, defines considering types of aggression (physical and verbal overt direct aggression), substantiates criteria for aggressive behavior assessment as a set of aggressive actions that define each aggression type. The methodology consists of the following stages: collecting video on the Internet, identifying time intervals where aggression is performed, localizing informants in video frames, transcribing informants’ speech, collective labeling of physical and verbal aggression actions by a group of annotators (raters), assessing the reliability of annotations agreement using Fleiss’ kappa coefficient. In order to evaluate the methodology a new audiovisual aggressive behavior in online streams corpus (AVABOS) was collected and labeled. The dataset contains audio and video segments that contains verbal and physical aggression correspondingly that manifested by Russian-speaking informants during online video streams. The results of interrater agreement reliability show substantial agreement for physical (κ = 0.74) and moderate agreement for verbal aggression (κ = 0.48) that substantiates the developed methodology. AVABOS dataset can be used in automatic aggression recognition tasks. The developed methodology can also be used for creating datasets with the other types of behavior.

Intelligent systems demand interaction with a variety of complex environments. For example, a robot might need to interact with complicated geometric structures in an environment. Accurate geometric reasoning is required to define the objects navigating the scene properly. 3D reconstruction is a complex problem that requires massive amounts of images. The paper proposes producing intelligent systems for 3D reconstruction from single 2D images. Propose a learnable reconstruction context that uses features to realize the synthesis. Proposed methods produce encoding feature lable input to classification, pulling out that information to make better decisions. Binary Classifier Neural Network (BCNN) classifies whether a point is inside or outside the object. The reconstruction system models an object 3D structure and learns feature filter parameters. The geometry and the corresponding features are implicitly updated based on the loss function. The training doesn’t require compressed supervision to visualize the task of reconstructed shapes and texture transfer. A point-set network flow results in BCNN having a comparable low memory footprint and is not restricted to specific classes for which templates are available. Accuracy measurements show that the model can extend the occupancy encoder by the generative model, which doesn’t request an image condition but can be trained unconditionally. The time required to train the model will have more neurons and weight parameters overfitting. 

Malware analysis is the process of dissecting malicious software to understand its functionality, behavior, and potential risks. Artificial Intelligence (AI) and deep learning are ushering in a new era of automated, intelligent, and adaptive malware analysis. This convergence of AI and deep learning promises to revolutionize the way cybersecurity professionals detect, analyze and respond to malware threats. This paper proposed a Deep Neural Network (DNN) model built from features selected by ANalysis Of Variance (ANOVA) F-test (DNN-ANOVA) to increase accuracy by identifying informative features. ANOVA is a feature selection method used for numerical input data when the target variable is categorical. The top k most relevant features are those whose score values are greater than a certain threshold equal to the ratio between the sum of all features scores and the total number of features. Experiments are conducted on CIC-MalMem-2022 dataset. Malware Analysis is performed using binary classification to detect the presence or absence of malware and multiclass classification to detect not only the malware but also its type. According to the test results, DNN-ANOVA model achieves best values of 100 %, 99.99 %, 99.99 %, and 99.98 % in terms of precision, accuracy, F1-score and recall respectively for binary classification. In addition, DNN-ANOVA outperforms the current works with an overall accuracy rate of 85.83 %, and 73.98 % for family attacks and individual attacks respectively in the case of multiclass classification.

Currently, switched reluctance motors are considered the most promising type of electromechanical energy converter without permanent magnets, especially for operations at sub-nominal speeds. To control of a switched reluctance motor to minimize torque ripple requires the regulation of phase currents based on the rotor angular position, utilizing the flux linkage as a function of both current and rotor angle. The flux linkage characteristic is essential in control systems that indirectly determine the rotor position. The paper presents an experimental methodology for deriving the flux linkage characteristic of a switched reluctance motor. The calculation of flux linkage for each rotor position angle of the electric machine is provided. The proposed methodology involves mechanically locking the rotor and periodically applying voltage to one of the motor phases using a power converter to gather data on phase current and voltage. Using the proposed experimental methodology, the relationships between flux linkage, phase current, and rotor angle were obtained. The results demonstrate that this methodology can be effectively utilized to accurately determine the flux linkage characteristic of a switched reluctance motor. The experimental methodology proposed in this paper can be employed to generate the flux linkage characteristic of a switched reluctance motor. This approach is particularly advantageous for designing model predictive control systems.

The article presents the results of the spectral dependence study of the quantum efficiency of photocatalytic water decomposition. The relationship between the radiation spectrum and the efficiency of photocatalytic water decomposition into hydrogen and oxygen is determined. For this purpose, an electrolyte based on sodium nitrate is studied. The photocathode contained nanoporous silver layers. It is shown that the maximum quantum efficiency of photocatalytic water decomposition by spectrum integrally amounts to 1.9 %, and increases with decreasing radiation wavelength. The obtained results can be used in the development of solar energy devices designed for photocatalytic water decomposition into hydrogen and oxygen.