ABSTRACT. The rapid development and progress of artificial intelligence algorithms in the last decade has opened up many new possibilities and fields for its application. The field of human computer interaction is not only not an exception, but it can also be considered a veteran. Among the general population, there is a certain level of apprehension when it comes to introducing artificial intelligence into new areas and aspects of human life. In this paper, we will attempt to shed new light on this issue and the current trend of villainizing AI, as well as present current trends in enhancing user security based on it. We will summarize the current conditions, trends and experiences in this area.

ABSTRACT. Traditionally, image recognition has been tackled as a classification problem. You only look once (YOLO) real-time object detection models take a different approach, treating recognition as a regression problem. Models utilize very efficient neural networks, making YOLO models suitable for real-time observation. Recent events have brought to light the importance of medical face masks in preventing the spread of highly contagious pathogens, as well as the extent individuals are willing to go to avoid wearing them. A robust model for detecting medical masks, capable of running in real-time could help improve the toolset available for preventing the spread of pathogens and reduce the impact contagious diseases could have in the future. This work compares the performance of several lightweight YOLO version eight models suitable for real-time application for mask detection. The advantages and shortcomings of various model architectures are assessed and their potential for use is explored.

ABSTRACT. Intense electromagnetic activity on the Sun's surface can lead to events known as solar flares that often lead to mass ejections and other solar events. Sufficiently powerful solar events can make it to the Earth and interfere with telecommunication systems. Notably energetic events can even affect satellite and ground communication infrastructure potentially causing massive damage. Extensive monitoring systems in combination with robust forecasting techniques can be used as an early warning system to minimize potential damage. This work proposes a method based on extreme gradient boosting (XGBoost) applied to solar flare classification. As the performance of XGBoost heavily relies on adequate hyperparameter selection, a modified metaheuristic algorithm is also introduced to tune network hyperparameters. The introduced approach has been evaluated on a real-world dataset and a comparative analysis has been carried out against several contemporary algorithms tackling the same task under identical conditions. The introduced approach has attained promising outcomes.

ABSTRACT. The shift towards renewable sources of power has brought several associated challenges. The sporadic nature of power generation, high reliance on weather, and difficulties with long-term storage are just some of the challenges facing large-scale adoption. Storage is especially challenging due to the large associated upfront costs of batteries coupled with limited lifetimes further shortened by uneven charging. One possible step for improving the adoption of renewable sources is to develop robust techniques for forecasting power generation to better plan power demands. This work proposed an approach based on time series forecasting through the use of recurrent neural networks augmented with an attention mechanism. To further optimize the performance of this approach a modified version of a well-known metaheuristic algorithm is also introduced. The proposed approach is evaluated on a real-world dataset consisting of weather and wind farm power generation data from a wind farm located in mainland China. Several metaheuristics have been applied to optimize the proposed approach and the outcomes have been subjected to a comparative analysis to determine the best approach.

ABSTRACT. Cloud load forecasting plays an important role in information technology. There is the increasing reliance on online services for accurately predicting service demand in order to take timely actions and guarantee better services and greater customer satisfaction. However, the complexities associated with making accurate forecasts make this a challenging task. By formulating this task as a time series forecasting problem powerful artificial intelligence algorithms able to handle temporal data such as the long short-term memory (LSTM) neural network may be used for this problem. Additionally, to help tackle the complex nature of the data, variational mode decomposition is incorporated into the approach. Better accuracy can be obtained by sleeting appropriate LSTM control parameters. Several well-known metaheuristics have been evaluated on their ability to optimize LSTM network parameters and improve performance when applied to a real-world cloud load forecasting dataset. The metaheuristic algorithms are subjected to comparative analysis to determine the best-performing approach.

ABSTRACT. This study focuses on analyzing historical data to forecast future trends in Bitcoin prices due to its influence on the business landscape. Its high volatility attracted attention to understanding the influencing factors for its price. This paper presents an empirical investigation using time-series data of various exogenous and endogenous variables. Closing prices of Bitcoin and Ethereum, along with the daily volume of Bitcoin-related tweets are examined for Bitcoin closing price prediction by a long-short term memory (LSTM) network, fine-tuned by a hybrid adaptive reptile search algorithm. The analysis covers a three-year period, in which data is divided into training, validation, and testing sets. Comparative analysis against LSTM networks tuned by other high-performing metaheuristic algorithms demonstrates that the novel approach outperforms competitors in terms of standard regression metrics.

ABSTRACT. This study conducts a comparative survey of some outstanding works on non-anthropomorphic robotic hands over the past four decades. In this research, the specific features of these works are identified by describing and evaluating their structures. In addition, the possibility of the practical application of these hands is provided. As a result, some comparative comments were also figured out. All these analyses and evaluations serve as the basis for the author’s design of the non-anthropomorphic hand. The paper concludes with a brief on the author's design.

ABSTRACT. In this paper, we consider rate redundancy not only of uniform scalar quantizer (USQ) for the assumed Laplacian source, but also of USQ coupled with Golomb-Rice coding. For a given bit rate of USQ, we select the key parameter of the Golomb-Rice coder so that the rate redundancy of USQ with Golomb-Rice coding is minimal. We further determine and analyze the robustness of the rate redundancy to changes in the standard deviation mismatch over a relatively wide range of values. The analysis presented in the paper is very useful, since in numerous systems with limited memory resources, any savings achieved by means of compression can indeed be beneficial.

ABSTRACT. The localization represents human and devices position determination. In fingerprinting technique, the machine learning algorithms are used to compare new and pre-existing fingerprints (location signatures) to estimate the required location. In this paper we analyzed the accuracy and confusion matrices of five machine learning algorithms to find the most suitable

ABSTRACT. The machine learning has a very broad specter of application in modern world. One of the important methods on which the machine learning relays is classification. Since the classification is very computational heavy process, the hardware accelerator can alleviate the work of processor and improve the machine learning process. In this paper we present our hardware accelerator which performs the k-NN classification.

ABSTRACT. The identification of the remote sensing images with non-monotonic dependence of the rate-distortion curve to the parameter that controls compression (strange images) is addressed in this research. Twenty-two simple features are extracted from the publicly available dataset with remote sensing images, and these features are compared using correlation between features and strange image membership function, as well as the correlations between analyzed features. Obtained results show that gray level entropy provides highest correlation with strange image membership function.

ABSTRACT. Rapid progress in machine learning has given rise to a multitude of approaches for various application domains. However, comparing the functionality and effectiveness of these approaches is crucial to obtain accurate and reliable results. In this study, we aim to explore the potential of ChatGPT, a cutting-edge conversational AI model, for coding applications. Our primary goal is to evaluate the effectiveness of ChatGPT in handling coding-related tasks. By conducting a systematic evaluation of the ten functions encompassed in our original dataset, we aim to gauge the effectiveness of ChatGPT in handling various coding tasks in case of Laravel Controller functions. This will enable us to critically examine the capabilities of ChatGPT in the context of our specific coding applications. Additionally, we conduct a survey among university students with a goal to assess adoption level of AI-based code assistance tools.

ABSTRACT. In this paper, we analyze the performance of wireless communication system affected by α-η-μ fading in the presence of α-η-μ co-channel interference (CCI) for the receiver with selection combining (SC) with L branches. Closed-form expressions for the system outage probability (Pout) of the ratio of the received signal and interference (SIR) are derived as a function of different values of the system parameters. Obtained results can be used in performance analysis of wireless communication system which uses diversity technique to reduce fading effects in fading channels. On the other side, in the second part of the paper, we investigate the potential of trending ChatGPT based on Large Language Mode (LLM) for task of QoS estimation, taking into account Pout among model inputs. Finally, the proposed method is compared to traditional machine learning algorithms using Weka in Java programming language.

ABSTRACT. The influence of the the mobile phone position on the electromagnetic field distribution inside the biological tissues of the user's head is discussed in the paper. The case of voice transmission was simulated at the frequency of 0.8 GHz. Numerical models of a child's head and a smartphone have been used. The distribution of the magnetic field within user’s head tissues for different positions of the mobile phone is considered.

ABSTRACT. In this paper, the assessment of the digital linearization technique’s influence on a dual-band power amplifier was carried out. The linearization technique requires utilization of the second-order nonlinear signals in the baseband which modulate the second harmonic of the carrier. The dual-band power amplifier (PA) operates concurrently at two widely spaced frequencies, 2.1 GHz and 5 GHz. The linearization of the dual-band is executed in simulation for both considered frequency bands independently. Results accomplished through the linearization process for various output power levels were acquired for FBMC, FOFDM and UFMC signals, which represent modulation forms for the 5G systems.

ABSTRACT. In this paper a new design of a double-slot antenna fed by microstrip line is proposed for power harvesting applications at Ka band. The antenna consists of two parallel slot radiators connected with CPW line in the middle. Radiation pattern of this antenna is optimized to exhibit the maximum gain of 9 dBi in two directions that are offset 40 degrees in respect to the antenna broadside. Proposed antenna is suitable for powering wireless sensor nodes and passive RF identification tags with known RF source when the frequency and direction of arrival of the field are known.

ABSTRACT. This paper studies possibilities of TLM method to analyze a textile slotted antenna for dual-band use. Different size of a polygon-shaped slot on rectangular patch antenna were analyzed to obtain dual band performance. Additionally, suitability of cylindrical TLM approach are demonstrated under bending conditions, where resonant frequencies and return loss for both bands were considered.

ABSTRACT. This paper presents a deep neural network (DNN)-based module designed for fast direction-of-arrival (DoA) estimation of a signal received by a multi-element textile wearable antenna array (TWAA). Two realizations of this module were considered with different levels of complexity of the artificial neural network structure, differing in the number of hidden layers of the neural network and in the total number of neurons in the hidden layers. In both cases, the neural networks were trained with both the classical full-batch approach using the Levenberg-Marquardt algorithm and the mini-batch approach using the RMSprop algorithm. An analysis of the influence of the complexity of the neural network structure and the choice of the algorithm for its training on the accuracy of the DoA module was performed. Also, all the results obtained with the DNN-based DoA module were compared with the DoA module implemented with the rootMUSIC algorithm.

ABSTRACT. In this paper, the cascade PNN-RBF neural model was developed to rapidly estimate the ground penetration depth of EM waves in the frequency range of 100-700 MHz. The input parameters of this model are soil composition (clay, silt and sand content), soil moisture and frequency. The model consists of a PNN network that determines the type of soil, pairs of RBF networks that determine the real and imaginary components of the relative permittivity of the soil based on the type of soil and its moisture, and function modules that estimate the penetration depth of EM waves based on the values of the complex relative permittivity and frequency. The effectiveness of this model is demonstrated by estimating the penetration depth of EM waves in clay loam soil.

ABSTRACT. Smart buildings promise to adapt environmental conditions to the needs of occupants based on statistical analytics applied to various monitored data. While sensors for accurate monitoring of building parameters such as temperature, lighting, and air-quality abound, sensors for occupant presence and comfort are wanting. Doppler radar sensors have shown great promise for unobtrusive recognition and monitoring of occupant presence, count, activity, and cardiopulmonary vital signs. With such measures, a smart building can optimize operations not only for the most efficient use of energy and space, but also to create healthy and sustainable environments that support occupant wellness, comfort, and productivity. This paper presents an overview of emerging sensors and systems redefining the concept of human-building interactive systems.

ABSTRACT. Good overall health depends on and affects one’s sleep quality. Insufficient or disordered sleep decreases cognitive function and predisposes patients to hypertension, stroke, and heart disease. An overnight sleep study can provide vital health diagnostics yet typically involves applying and monitoring multiple body-contact sensors, which can interfere with sleep and require cumbersome manual data analysis. Doppler radar technology has been demonstrated to provide a non-invasive means of measuring vital signs through clothing and bedding, including respiratory rate, heart rate, activity, body position, and tidal respiratory volume. This paper examines the potential for applying physiological radar to assess sleep apnea in a manner suitable for aiding patients and their caregivers in developing effective intervention strategies.

ABSTRACT. With the increased interest of non-contact vital sign detection, the need for a small, compact radar has emerged. The development of millimeter-wave radars and Antenna-on-Package (AoP) solutions provides a solution for a portable, small form factor device. In this paper, a commercial, compact millimeter wave radar operating at 60 GHz is used to accurately detect vital signs. With the use of a super-resolution direction of arrival algorithm, digital beamforming, and a frequency carving algorithm, various experiments, involving multi-subjects and drywall, are performed to demonstrate the accuracy and versatility of the radar system.

ABSTRACT. As an important physiological signal, the electrocardiogram (ECG) is wildly used to reflect human heart activity, which is essential for disease detection and life safety protection. However, traditional ECG monitoring means rely on contact electrode pads, which are costly and user-unfriendly. Inspired by the success of millimeter wave radar for close-range motion sensing, we propose a novel deep learning method to synthesize periodic ECG signals from millimeter wave radar signals to achieve contactless real-time ECG monitoring. Specifically, we use the bidirectional long short-term memory network to fit the mapping relationship between radar signals and real ECG, which is trained end-to-end. Experimental results on a real-world dataset show that the proposed method can effectively synthesize ECG signals for the RF signal capture by the millimeter wave radar and achieve accurate ECG monitoring contactlessly. In addition, the well-trained model maintains well generalization to the real-world millimeter wave radar signals containing various random perturbations.

ABSTRACT. Unmanned aerial vehicle (UAV) is an emerging and promising technology used in various fields and applications. Its usage has grown in recent years since they are becoming more accessible to users, and more appealing applications are constantly being developed. This paper presents a drone detection and tracking solution using a frequency-modulated continuous-wave (FMCW) radar with multiple-input multiple-output (MIMO) capability that analyzes a drone performing the pitch, roll, yaw, and throttle variation motions. The movements are detected by analyzing the micro-Doppler (mD) signatures associated with the drone, and since MIMO enables digital beamforming capability by taking advantage of an antenna array, azimuth, and elevation angles were also analyzed. By looking at the angle of arrival of the drone signal, it is possible to keep track of the UAV and differentiate its motions. Compared with existing works, this paper focuses on analyzing and differentiating the main drone motions performed during flight using a portable FMCW MIMO radar with digital beamforming that follows the drone’s movements in azimuth and elevation directions.

ABSTRACT. This contribution describes the preliminary project of a radar system designed to detect space debris. Whereas conventional debris detection systems rely on ground-based radar or optical instruments, the proposed radar is targeted to be mounted on a satellite. As a consequence of the space environment peculiarities, several technical constraints arise in addition to the requirements concerning the classic target detection. The technical characteristics of the radar payload together with the expected performance will be discussed throughout the text.