One of the most important problems in digital speech-signal processing is distinguishing segments of active speech and of background noise or silence in an input acoustic signal. This problem arises in many important practical applications, such as speech analysis in voice command systems, transmission of speech over a network, automated speech recognition, etc. However, most available systems designed for automated speech analysis cannot efficiently solve this problem if the signal-to-noise ratio is small. In addition, their parameters must be tuned separately for different noise levels. This prevents fully automated segmentation of noisy speech signals. In this work, we design a system for the automated segmentation of speech signals distorted by additive noise of different types and intensities. The developed system is based on three various deep convolutional neural network models and can efficiently detect speech and silence segments in noisy signals over a wide range of the signal-to-noise ratios and different noise types.

An alternative approach to speech denoising using generative diffusion models that model the distribution of training data is proposed. In recent years, such models have led to promising results to be obtained in the field of generating signals of various kinds, and these are superior in many ways to previous generative models, such as variational autoencoders. However, diffusion models have not yet found wide application in the field of speech denoising. A new diffusion model is presented, which can be used to denoise real speech signals using a deep neural network. Our own data set, with more than 150 h of pure speech in Russian, has been created. The obtained results, estimated using the metrics scale invariant signal to distortion ratio and perceptual evaluation of speech quality, are comparable or superior to the results of the best discriminative models.

In this paper, a multiclass image semantic segmentation problem was solved. For analysis, images of the intracytoplasmic sperm injection process were used. For training the neural network, 656 frames were manually labelled. As a result, each pixel of the images was assigned to one of four classes: microinjector, suction micropipette, oolemma, background. An analysis of modern approaches was carried out and the best architecture, encoders, and hyperparameters of the neural network were selected experimentally: the convolutional neural network FPN (feature pyramid network) with the resnext101 encoder having a depth of 101 layers with 32 parallel separable convolutions. The developed neural network model has allowed obtaining the segmentation efficiency of IOU=0.96 at the algorithm speed of 15 frames per second.

In this paper we describe the problem of painting style classification into five classes: impressionism, realism, expressionism, post-impressionism and romanticism. While most previous approaches relied on image processing and manual feature extraction from painting images, our model based on the ResNet architecture and pre-trained on the ImageNet dataset operates on the raw pixel level. The training has been performed on a large dataset (about 43k images for five class style classification problem). To increase the quality of final model a large number of various augmentations were used: random Affine transform, crop, flip, color jitter (i.e. contrast, hue, saturation), normalization, a scheduler for the optimizer. Finally model weights were pruned which allowed increasing accuracy up to 51.5% and decreasing computation time as well.

A possibility of assessing the electromagnetic emission of stationary plasma thrusters (SPTs) using standard metal vacuum chambers considering the origination of the natural resonances in the latter is studied. The primary wave modes that originate in cylindrical metal test chambers are defined and their influence on the measurement accuracy is assessed. To explain the nature of the SPT natural oscillations, the authors propose and analyze a model based on an equivalent coaxial resonator that describes resonant properties of the SPT gas discharge chamber. Analysis is made for the emission spectra of SPTs of various models. The analysis revealed coincidence of the spectrum components in frequency with the natural oscillations of the resonator that simulates the SPT gas discharge chamber.

A significant step towards the comprehensive solution of the important new fundamental control problems of high-temperature plasma of toroidal configuration is made by utilizing heterogeneous mini-supercomputers of NIISI RAS (Scientific Research Institute of System Development, Russian Academy of Science). The basic components of hardware and software modeling system and automatic real-time plasma feedback control are developed and implemented. The system is named HASP CS (Hardware-Software Plasma Control System). It is based on a fundamentally new, more accurate and reliable control algorithm using a solution with a given accuracy of a number of inverse diagnostic problems by epsilon-nets. The possibility of an application of the proposed technique during discharge time to provide feedback control of the boundary and internal plasma parameters is demonstrated. It is shown that the complex allows investigating the processes in currently operated and designed fusion devices and solving problems of fusion application for energy production. The recommendations to the engineering implementation of plasma control systems are made. The directions for further research are formulated.

A number of modern technologies have been applied for the organization of distributed access to the informational and computational resources of the plasma modeling and control complex HASP CS. These allowed to ensure correct configurations of HASP CS on different computational platforms, effective work of users and a sufficient level of security. The Docker technology is used for creating isolated software environments and jobs. The NoVNC technique is applied for remote desktop access. The software structure of the HASP CS complex is described. The main difficulties and some optimal solutions of the distributed access to the complex science-intensive computer systems are discussed.

A procedure for analysis, calculation, and optimization of the parameters of electric system for power supply of inductive discharge is developed. The conditions for the maximum efficiency of the power supply systems in high-frequency ion engines are analyzed. An equivalent circuit of the inductive discharge with capacitive coupling is proposed using the analysis of schemes of different high-frequency ion engines and transformed into a complex equivalent impedance connected in the inductor circuit to calculate matching reactive elements. Equivalent parameters of the plasma of inductive discharge are estimated using the wave model. Calculated and experimental results are used to formulate the main requirements on the laboratory models of the high-frequency oscillators and matching units of the radio-frequency ion engines.

The analysis for the optimal energy input from the radio frequency (RF) generator into the inductive discharge is carried out. The analysis was performed using the developed equivalent circuits of the inductor with the inductive discharge plasma of the radiofrequency ion thruster. Taking into account the wave model of radiofrequency discharge it was shown that due to the plasma reaction the inductance of coil may vary several times vs. the geometry of the discharge chamber and the plasma density. The mathematical and numerical models of the discharge electric power supply system have been developed and the results of calculations for some models of radiofrequency ion thrusters are presented. In particular, maximum efficiency of the energy input into the considered RF ion thruster discharges was estimated.

Wireless modeling is an extremely important part of wireless networks research and design. Nowadays many advanced of numerical (simulation) codes are being developed and widely used in different areas of wireless networks research. The key problem is that such codes are often proprietary and poorly documented. Another problem is that most codes cannot be used together due to different input/output formats. A set of data adapters and converters should be developed and adopted to couple such codes. At the same time, a number of problems with data processing, monitoring and results visualization must be solved. "virtual computational network" (VCN) modeling environment is designed as a universal easy-to-use toolbox to create complicated modeling cases and scenarios for numerical experiments built atop of different codes and pre-calculated data. VCN is a powerful tool supporting distributed parallel computing, input/output data transformation, and results visualization "out-of-the-box", allowing end-users to greatly improve numerical modeling studies efficiency.

Applications possibilities of water, ethanol and water-ethanol conductometric biomedical sensors are discussed as an alternative to existing measuring instruments of electromagnetic radiation and concentration of carbon dioxide generated through the skin of a person. The causes of anomalous increase of electrical conductivity in water and ethanol discovered in our experiments are considered. It is shown that the phenomenon is attributed to the peculiarities of CO2 absorption in liquids and of radiation spectra of human skin and absorption spectra of water and ethanol in infrared and THz frequency ranges. Influence rate of CO2 on water conductivity is estimated. The possible nature of observed phenomena and sensor features are discussed. Experimental and theoretical results are in a good agreement.

Nowadays the gas discharges are well-studied and there are a number of application codes for plasma simulation. The key problem is that such codes are often proprietary and poorly documented. Another problem is that such codes cannot be used together due to different input/output data formats. If some plasma simulation code is not an open source one, the only way to reuse its numerical results is to write an application or script which will convert the output data to the proper format. “Virtual Discharge” modeling environment was developed to allow users to build complicated modeling cases only with several mouse clicks by the help of simply-to-use graphical user interface. Having a chance to configure different code calculation chains and to define data convertors, a user can easily reuse third-party plasma simulation codes saving his time. “Virtual Discharge” is developed with Java so it’s a cross-platform application. Being integrated with ScopeShell data analysis and visualization integrated shell [1] and Tadisys task distribution system [2] it is a powerful tool for complicated plasma simulation cases, which also supports distributed computations.

The results of application of automatic control methods to the problem of the given total current sustainment in toroidal plasma are presented. Plasma current control is performed mainly by solenoid with the use of electromagnetic inductance effect. The mathematical problem is formulated, the methods and developed software are described, and the solutions of test and real problems are given. The comparison of experimental and numerical data allows drawing a conclusion on the nature of plasma conductivity in spherical tokamaks.