2687-0517
Computing, Telecommunication and Control
18
3
2025
1-153
RAR
RUS
9-22
0000-0001-9181-7726
Moscow Technical University of Communications and Informatics; Moscow Aviation Institute (National Research University)
Agamirov
Vladimir
avhere@yandex.ru
Moscow, Russian Federation
0009-0009-6909-9399
National Research University “Moscow Power Engineering Institute”; Moscow Technical University of Communications and Informatics
Agamirov
Levon
itno_agamirov@mail.ru
Moscow, Russian Federation
BRICS University
Nosikov
Maksim
nosikovmaxim@yandex.ru
Moscow, Russian Federation
0000-0002-2851-8472
Moscow Technical University of Communications and Informatics
Toutova
Natalia
e-natasha@mail.ru
Moscow, Russian Federation
0009-0008-7929-5492
Moscow Technical University of Communications and Informatics
Khaush
Anton
anton.khaush@gmail.com
Moscow, Russian Federation
Application of neural networks for detecting defects and damage in metal structures
The rapid development of neural networks has led to the integration of these technologies into various industrial sectors. At the same time, improving the accuracy and efficiency of detecting defects and damages, including in real-time, remains a critical task. By combining neural networks with the Internet of Things (IoT) and technologies for data collection, storage and protection, it is possible to create a comprehensive and effective information-measurement system for surface defect detection. In this context, the present work highlights recent advances in the application of artificial intelligence for quality control, as well as the detection of defects and damages in structures. The focus is on the development and training of neural networks capable of effectively identifying and classifying various types of defects. The study demonstrates how these technologies significantly improve the speed and accuracy of diagnostics compared to traditional visual and instrumental inspection methods. The results of model testing on real industrial data confirm the high efficiency of the proposed approach. Additionally, the authors have developed an algorithm and implemented software for the automatic annotation of images in a format suitable for modern architectures such as YOLO. This approach enables the effective application of the model for detecting damages on the surfaces of structures and systems using widely available types of datasets.
10.18721/JCSTCS.18301
620.179; 004.93
defect detection
neural networks
YOLO architecture
structural integrity monitoring
infrared thermography
automated defect analysis
machine learning for defectoscopy
https://infocom.spbstu.ru/article/2025.86.1/
RAR
RUS
23-35
Peter the Great St. Petersburg Polytechnic University
Mbele Ossiyi
L.P.
lucprucell@gmail.com
St. Petersburg, Russian Federation
56049610600
0000-0003-1116-7765
Peter the Great St. Petersburg Polytechnic University
Drobintsev
Pavel
drobintsev_pd@spbstu.ru
St. Petersburg, Russian Federation
6603839750
Peter the Great St.Petersburg Polytechnic University
Sergey M. Ustinov
usm50@yandex.ru
Polytechnicheskaya, 29, St.Petersburg, 195251, Russia
Application of machine learning algorithms and neural networks for analyzing the influence of data type in hate speech detection
At present, communication has reached an unprecedented level of activity thanks to online social platforms that have overcome geographical and linguistic barriers. However, the shift to online communication is accompanied by the spread of hate speech, which negatively affects the social environment of these platforms. In the field of natural language processing, research is being conducted to develop models for detecting and classifying hate speech, aimed at improving the safety and quality of the online environment. However, many of these studies are based on commonly used datasets that turn out to be unbalanced and insufficiently adapted to the new grammatical features of hate speech. This article presents a comparative study of the effectiveness of machine and deep learning algorithms in detecting hate speech based on a synthetic dataset. Three separate experiments were conducted using original and synthetically perturbated data. The findings indicate that employing a synthetic dataset enhances the representation of extremely negative or infrequently encountered communication scenarios, contributing to their more effective detection. Deep learning algorithms demonstrated superior performance in all experiments. The top-performing models in the first and second experiments, both using zero-shot learning, yielded accuracies of 52.04% and 62.13%, respectively. The last experiment revealed that the BiGRU + fastText architecture outperformed other models, achieving an accuracy of 72.68%.
10.18721/JCSTCS.18302
004.8
sentiment analysis
emotion recognition in text
attention mechanism
embedding
CNN
LSTM
GRU
https://infocom.spbstu.ru/article/2025.86.2/
RAR
RUS
36-45
Russian State Social University (RSSU)
Turchinskii
Kirill
turchin.sky@yandex.ru
Moscow, Russian Federation
Russian State Social University (RSSU)
Krasnov
Andrey
krasnovmgutu@yandex.ru
Moscow, Russian Federation
Automation of biological cell image processing
When analyzing images of biological cells, automated methods for segmentation and result storage are becoming increasingly in demand. Manual annotation is extremely labor-intensive and does not scale to large volumes of data, while conventional segmentation algorithms create binary masks of substantial size. The objective of this work is to develop a software pipeline that combines local threshold filtering and morphological post-processing to obtain an accurate binary mask and then encodes the result using run-length encoding (RLE) to reduce storage space. Methods used are as follows: at the segmentation stage, local statistical criteria are applied, followed by morphological closing. For storing the result, several modifications of RLE (standard, Foreground-Only, DRLE and Z-order) are implemented along with their comparative analysis. The scientific novelty of the work lies in the comprehensive integration of block filtering and morphology with subsequent compression of binary segmentation masks in the task of erythrocyte (and other cells) segmentation. This approach significantly reduces storage requirements without substantial loss of accuracy. The proposed solution demonstrates high metrics (Accuracy, IoU, Dice) while substantial memory savings. The practical significance is that the developed software pipeline can be easily integrated into biomedical data analysis systems, accelerating the mass processing of cell images and reducing the demands on storage infrastructure.
10.18721/JCSTCS.18303
004.932
segmentation
biological images
run-length encoding
local threshold filtering
morphological post-processing
automation
accuracy
https://infocom.spbstu.ru/article/2025.86.3/
RAR
RUS
46-57
Peter the Great St. Petersburg Polytechnic University
Aghayev
Aslan
agaev.af@edu.spbstu.ru
St. Petersburg, Russian Federation
Molodyakov
Sergey
sm50@yandex.ru
6603839750
Peter the Great St.Petersburg Polytechnic University
Sergey M. Ustinov
usm50@yandex.ru
Polytechnicheskaya, 29, St.Petersburg, 195251, Russia
Text augmentation method via paraphrastic concept embeddings: A case study on Azerbaijani language
A novel data augmentation method – paraphrastic concept embeddings – is presented, designed to address the problem of insufficient labeled data in Azerbaijani natural language processing (NLP). This method generates high-quality paraphrastic sentences by encoding semantic concepts into a continuous vector space and decoding them into diverse textual realizations. This approach is the first to utilize concept-level paraphrasing for the Azerbaijani language, yielding substantial improvements in applied tasks. The theoretical foundations of the method, including its mathematical formulation and implementation within NLP pipelines, are proposed. In text classification experiments, the method outperforms standard augmentation techniques in accuracy and robustness. The method does not require external lexical resources, making it especially useful for low-resource languages. It scales for various types of tasks, including sentiment analysis, entity extraction and text generation. It is concluded that the proposed approach significantly advances the level of Azerbaijani NLP and has the potential to be extended to other low-resource languages.
10.18721/JCSTCS.18304
004.89
natural language processing
low-resource language
data augmentation
paraphrastic embeddings
concept embedding
text classification
https://infocom.spbstu.ru/article/2025.86.4/
RAR
RUS
58-67
0000-0002-7437-6153
Pavlov
Evgeniy
pavlov_ea@spbstu.ru
56049610600
0000-0003-1116-7765
Peter the Great St. Petersburg Polytechnic University
Drobintsev
Pavel
drobintsev_pd@spbstu.ru
St. Petersburg, Russian Federation
Peter the Great St. Petersburg Polytechnic University
Klinkov
Victor
klinkovvictor@yandex.ru
St. Petersburg, Russian Federation
Peter the Great St. Petersburg Polytechnic University
Semencha
Alexander
asemencha@spbstu.ru
St. Petersburg, Russian Federation
Peter the Great St.Petersburg Polytechnic University
Igor
G.
igcher@spbstu.ru
Polytechnicheskaya, 29, St.Petersburg, 195251, Russia
Method for automated enrichment of a knowledge base on glass compositions and properties based on data from scientific publications
Automating the extraction of glass composition and property data from scientific literature is critically important for accelerating the development of new material. This work presents a method integrating: 1) the collection of full-text articles using the Elsevier Research Products APIs, 2) text preprocessing, 3) context-dependent extraction of structured data using a large language model (LLM) and a domain-specific prompt, 4) enrichment of a knowledge base on glasses. The key achievement is the development of a prompt that yields an F1-score of 0.99 for extracting chemical compositions, their properties and correctly establishing relationships between them on a sample of 50 articles. The proposed method significantly simplifies the automatic creation and continuous updating of knowledge bases on glasses, thereby eliminating the traditional reliance on manually curated, potentially outdated resources and providing a robust, data-driven foundation for the efficient designing of glasses with target properties using machine learning.
10.18721/JCSTCS.18305
004.89
data extraction
natural language processing
LLM
prompt engineering
knowledge base
glass
glass properties
https://infocom.spbstu.ru/article/2025.86.5/
RAR
RUS
68-79
Peter the Great St. Petersburg Polytechnic University
Pyatlin
Artem
poccomaxa@cave3d.com
St. Petersburg, Russian Federation
Morozov
Dmitriy
dvmorozov@inbox.ru
Design of filters using pseudo resistors for biomedical devices
The current state of the scientific and technical problem being solved is assessed, initial data are obtained, a 4th-order Sallen–Key filter and a filter with imitation of the inductors by an active circuit based on impedance converters using operational amplifiers and pseudo resistors are developed, and the results are compared. It is recommended to use the low-pass filter with imitation of the inductors in an electronic stethoscope, since it has, compared to the Sallen–Key filter, a less roll-off in the passband, greater attenuation in the stopband, a sharper drop in the frequency response in the transition region, better noise characteristics and a larger dynamic range. The filter with imitation of the inductors has lower nonlinear distortions and demonstrates operability with a spread of temperatures and element ratings, especially in the frequency range containing the main peaks of heartbeat and lung sounds, and the power consumption and hardware costs of such a filter are comparable to similar characteristics of the Sallen–Key filter.
10.18721/JCSTCS.18306
621.3.049.774.2
electronic stethoscope
Sallen–Key filter
imitation of inductance
pseudo resistor
negative impedance converter
noise characteristic
integrated circuit layout
https://infocom.spbstu.ru/article/2025.86.6/
RAR
RUS
80-88
0009-0006-6987-7144
“Special Technological Center” JSC
Ivannikova
Victoria
ivannikova.vicky@yandex.ru
St. Petersburg, Russian Federation
“Special Technological Center” JSC
Korotkov
Vladimir
diofant2912@gmail.com
St. Petersburg, Russian Federation
Algorithm for automatic recognizing of radar scan type, based on the extraction of statical features from input analyzed process
An algorithm for recognizing the radar antenna scan type by radio monitoring station is proposed. It is based on the extraction of a set of statistical features from the input radio signal, characterizing the signal amplitude envelope, modulated according to the scan type of radar antenna pattern. For feature extraction, mechanical, one-dimensional electronic and two-dimensional electronic scanning are considered, along with the beam steering schemes of radar antenna pattern used in practice. The essence of the proposed algorithm is the sequential comparison of the features with certain thresholds. The proposed recognition scheme is relevant for tasks of classifying the radio electronic equipment signals in a complex radio electronic environment and resolving ambiguity in recognizing radio signal sources with overlapping regions in the feature space, which characterizes time-frequency parameters of the radio signal. Experimental results are given, showing that the proposed algorithm ensures high recognition quality and is promising for improving existing and developing new radio monitoring equipment.
10.18721/JCSTCS.18307
621.396.96
radio emission source recognition
radar systems
antenna pattern
radar scan type
electronic scanning
mechanical scanning
https://infocom.spbstu.ru/article/2025.86.7/
RAR
RUS
89-101
0009-0004-1628-1772
Peter the Great St. Petersburg Polytechnic University
Pham
Huu Duc
phamduc2511997@gmail.com
St. Petersburg, Russian Federation
22735712200
Peter the Great St. Petersburg Polytechnic University
Sorotsky
Vladimir
sorotsky@mail.spbstu.ru
St. Petersburg, Russian Federation
0000-0001-9738-9291
Peter the Great St. Petersburg Polytechnic University
Zudov
Roman
rzudov@spbstu.ru
St. Petersburg, Russian Federation
Reduced voltage stress Class E power amplifier operating a complex impedance load: A performance analysis
Class E power amplifiers (PAs) attract the interest of experts involved in the development of communication and telecommunications equipment due to their high efficiency. However, the high voltage stress across transistors, which exceeds the supply voltage by 3.6–4 times, limits the output power of such amplifiers. An alternative to solve this issue could be a PA in which the peak voltage across transistor is reduced by 2 times, but still maintains the main advantages of traditional Class E, such as zero-voltage switching (ZVS) and zero-derivative voltage switching (ZDVS). In well-known publications, the study of the characteristics of PAs with lower voltage across transistors is limited to the particular case of a real impedance load. However, this condition may not be true when the PA operates in a frequency band, which will inevitably lead to errors in calculating their characteristics. The objective of the paper is to develop an analytical model of Class E PA with reduced voltage stress when operating with complex impedance load. The adequacy of the analytical model is confirmed by simulation, which shows that the relative error in the calculation of the main characteristics of the PA does not exceed 6.5%. The issues of synthesizing a filtering and matching circuit have been considered that ensures expansion of the frequency band at specified rated values in output power and voltage stress in transistor turned-on moment.
10.18721/JCSTCS.18308
621.37
power amplifier
сlass E
switching power losses
complex impedance load
harmonic balance method
https://infocom.spbstu.ru/article/2025.86.8/
RAR
RUS
102-110
The Bonch-Bruevich Saint-Petersburg State University of Telecommunications
Filin
Vladimir
filin_vladimir@mail.ru
22 Prospekt Bolshevikov, St. Petersburg, 193232, Russian Federation
0000-0001-6942-0618
Tashkent Universirty of information technology named after Muhammad al-Khwarizmi, Samarkand branch
Sattarov
Khurshid
s.xurshid@tuit.ru
Samarkand, Uzbekistan
0000-0001-6317-8007
Bonch-Bruevich St. Petersburg State University of Telecommunications; North-Western State Medical University named after I.I. Mechnikov
Yurova
Valentina
va-yurova@mail.ru
St. Petersburg, Russian Federation
Bonch-Bruevich St. Petersburg State University of Telecommunications
Golovin
Alexei
cathseugut@yandex.ru
St. Petersburg, Russian Federation
Selection of the optimal structure of a transformer based on single-turn elements for high-power switching transistor harmonic oscillators
Most electronic devices use power supplies and signal generators. A rationally designed high-power switching radio frequency (RF) generator usually contains one oscillatory circuit, a transformer and a grounded load. This work presents the rationale and calculations for the proposed structure of an RF transformer for high-power switching harmonic oscillators, based on the connection of unit elements. An analysis of the obtained calculation results was carried out to select the most optimal structure for most practical applications. It has been established that using N single-turn transformers achieves an N-fold gain in the volume and weight of ferrite and copper tubing. However, this gain is not always convenient to implement due to design considerations. The presented calculations confirm that the proposed structure is the most optimal for most practical applications, as it is free from many drawbacks of the classical high-power transformer scheme with multi-turn windings.
10.18721/JCSTCS.18309
621.375.026
semiconductor electronics
transformer
switching HF generator
power supply
electronic components
https://infocom.spbstu.ru/article/2025.86.9/
RAR
RUS
111-122
Peter the Great St. Petersburg Polytechnic University
Lomsadze
Dzhuletta
dlomsadze31@yandex.ru
St. Petersburg, Russian Federation
0000-0003-2595-4903
Peter the Great St. Petersburg Polytechnic University
Kolmakova
Natalia
kolmakova.nataliya@gmail.com
St. Petersburg, Russian Federation
0000-0001-7726-8492
Volvenko
Sergey
Topology and design methodology of a quadrature waveguide power divider for an arbitrary frequency range
This article focuses on the modeling of a quadrature waveguide power divider, taking into account technological requirements and manufacturing feasibility. A relatively simple yet
effective device topology is proposed, along with a method for calculating its geometric parameters to operate reliably within a predefined frequency range. The study includes a comprehensive analysis of existing power divider configurations, provides a detailed overview of the theoretical foundations for designing a quadrature waveguide-slot bridge and an in-depth examination of all design stages of the power divider, using the 27–35 GHz frequency range as an example. Based on the developed model, a working prototype was fabricated. Measurements of its frequency characteristics confirmed the validity and practical applicability of the proposed approach. The design methodology developed in this work can be effectively applied to create a quadrature waveguide-slot bridge in any given frequency range, making it a versatile tool for engineers and researchers in the field of microwave technologies.
10.18721/JCSTCS.18310
621.372.8
quadrature power divider
waveguide-slot bridge
rectangular waveguide
calculation method
device topology
https://infocom.spbstu.ru/article/2025.86.10/
SCO
RUS
123-130
Peter the Great St. Petersburg Polytechnic University
Bystrov
Vitaly
bystrov.vd@edu.spbstu.ru
St. Petersburg, Russian Federation
Morozov
Dmitriy
dvmorozov@inbox.ru
Pilipko
M.M.
m_m_pilipko@rambler.ru
An 8-bit wide input swing analog-to-digital converter based on voltage-controlled oscillator
An 8-bit analog-to-digital converter based on a voltage-controlled oscillator using 180 nm CMOS technology from Mikron JSC with a supply voltage of 3.3 V for the analog part
and 1.8 V for the digital part is presented. The analog-to-digital converter has a wide range of input voltages from 0 to 3.3 V. The transistor-level simulation of the analog-to-digital converter in the time domain was performed in Cadence Virtuoso. The sampling rate was set to 1 MHz. The power consumption is about 1.8 mW. The dimensions of the designed layout are 103 μm by 109 μm. With an input frequency of 50 kHz and an amplitude of 1.55 V, the post-layout simulation shows an output SNDR of 36.48 dB (ENOB is 5.77 bits).
10.18721/JCSTCS.18311
621.3.049.774.2
analog-to-digital converter
voltage-controlled oscillator
linearity of the transfer characteristic
digital synthesis
https://infocom.spbstu.ru/article/2025.86.11/
RAR
RUS
131-143
0009-0004-8263-3289
Peter the Great St. Petersburg Polytechnic University
Kiseleva
Aleksandra
aleksandrakiseleva2001@gmail.com
St. Petersburg, Russian Federation
Nikiforov
Igor
igor.nikiforov@gmail.com
0009-0005-3406-7005
Peter the Great St. Petersburg Polytechnic University
Shishko
Maxim
shishko.mv.post@gmail.com
St. Petersburg, Russian Federation
6603839750
Peter the Great St.Petersburg Polytechnic University
Sergey M. Ustinov
usm50@yandex.ru
Polytechnicheskaya, 29, St.Petersburg, 195251, Russia
Automation of preparation and deployment of information infrastructure of cloud services using the Ansible tool
The preparation and deployment of information technology infrastructure when implementing corporate collaboration platforms in industrial companies takes up to 40% of total time spent on a project. Automation of these processes helps reduce indicated time costs. This article proposes an approach to automate the preparation and deployment of information infrastructure, as well as to automate the installation of cloud services using the Ansible application. The R7-office platform is chosen as the implemented solution, which has basic necessary functionality for working with documents of various formats, enables collaborative document editing and meets state requirements for import substitution of components. The approach is based on an algorithm for preparing information technology infrastructure and installing the target platform using a software tool written in the Python programming language. It enables automatic parsing of formalized requirements, generating access rights requests for users and creating playbooks for configuring the infrastructure using Ansible. According to the experimental results, the implementation of the proposed approach reduces the labor intensity of the infrastructure preparation process by 45%.
10.18721/JCSTCS.18312
004.896
R7-office
collaborative editing
automation
IT infrastructure
Ansible
https://infocom.spbstu.ru/article/2025.86.12/
RAR
RUS
144-153
Sazanov
Arseniy
arseny.sazanov@gmail.com
AAH-8784-2019
35303230700
Peter the Great St.Petersburg Polytechnic University
Vyacheslav
P.
shkodyrev@imop.spbstu.ru
Polytechnicheskaya, 29, St.Petersburg, 195251, Russia
6603839750
Peter the Great St.Petersburg Polytechnic University
Sergey M. Ustinov
usm50@yandex.ru
Polytechnicheskaya, 29, St.Petersburg, 195251, Russia
Development of a dual-loop method of intelligent traffic light control based on reinforcement learning and hourly distillation of phase strategies
With increasingly complex urban dynamics, as well as increasing demands for the sustainability of urban mobility and introduction of cognitive technologies into transport infrastructure, the paper proposes a dual-loop method for intelligent traffic light control based on
reinforcement learning and phase strategy distillation procedures. The first level implements real-time control through an RL-agent, while the second one generates backup hourly plans based on statistics of its behavior. The method is based on a system-discrete model taking into account stochastic traffic parameters and permissible control constraints. The simulation conducted in SUMO for a real intersection demonstrates a significant reduction in average transport delay compared to classical control, confirming the efficiency, sustainability and scalability of the approach. The obtained results substantiate the possibility of practical implementation of the model within the framework of intelligent transport systems of large cities and for laying the engineering foundation for hybrid urban mobility management architectures.
10.18721/JCSTCS.18313
004.021
reinforcement learning
intelligent traffic light control
dual-loop control architecture
traffic light controller
traffic management and control
https://infocom.spbstu.ru/article/2025.86.13/