Alexander Dvorkovich
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Alexander Dvorkovich has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Rui Fan
College of Electronics and Information Engineering, Tongji University, Shanghai, China
6 shared publications - Liubov Antiufrieva
MIPT, Moscow, Russia
4 shared publications - Konstantin Iansitov
MIPT, Moscow, Russia
3 shared publications - Aleksandr Ivchenko
Keldysh Institute of Applied Mathematics of RAS
2 shared publications - Cairong Zhao
School of Computer Science and Technology, Tongji University, Shanghai, China
2 shared publications - Chuang-Wei Liu
College of Electronic and Information Engineering, Shanghai Institute of Intelligent Science and Technology, Shanghai Research Institute for Intelligent Autonomous Systems, Shanghai Key Laboratory of Intelligent Autonomous Systems, State Key Laboratory of Autonomous Intelligent Unmanned Systems, and Frontiers Science Center for Intelligent Autonomous Systems (Ministry of Education), Tongji University, Shanghai, China
2 shared publications
Latest Publications
- 29 April 2026
- Published by Bonch-Bruevich State University of Telecommunications in Proceedings of Telecommunication Universities
- Vol. 12 (2)
Energy Efficiency in Multi-Cell Massive MIMO: Impact of Signal Processing Schemes and Hardware Implementation Parameters
This paper investigates the trade-off between energy efficiency and spectral efficiency in multi-cell massive MIMO systems. The relevance of the study is driven by the need to simultaneously increase throughput and reduce the energy consumption of base stations in next-generation wireless networks, given the growing number of antennas and served users. The purpose of the study is to determine how the number of base-station antennas, the number of served users, linear signal processing schemes, and hardware implementation parameters affect the energy efficiency of multi-cell massive MIMO systems, and to identify configurations that provide the best trade-off between energy and spectral efficiency. Methods. A mathematical model of a massive MIMO system is developed, taking into account the number of base station antennas M, the number of user equipment K, and various linear signal processing schemes, including MR, ZF, RZF, S-MMSE, and M-MMSE. The model incorporates power consumption parameters reflecting hardware implementation characteristics, represented by two different sets of component specifications. The evaluation of energy and spectral efficiency is carried out using simulation-based analysis for various system configurations. Results. The results show that an optimal antenna-to-user ratio of M/K ≈ 3–4 achieves maximum energy efficiency without a significant reduction in spectral efficiency. It is demonstrated that the M-MMSE and S-MMSE algorithms provide the highest energy efficiency performance with moderate computational complexity, particularly when improved hardware components are employed. The obtained results confirm the existence of a pronounced energy efficiency optimum as the number of base station antennas increases . The novelty of this work lies in the comprehensive analysis of the energy–spectral efficiency trade-off in multi-cell massive MIMO systems while jointly accounting for linear signal processing schemes and hardware implementation parameters, which enables the formulation of practical recommendations for base station configuration under technological constraints. Practical significance. The findings of this study can be applied to the design and optimization of energy-efficient multi-cell massive MIMO systems for next-generation wireless communication networks, taking into account hardware implementation constraints and quality-of-service requirements.
- 29 April 2026
- Published by Bonch-Bruevich State University of Telecommunications in Proceedings of Telecommunication Universities
- Vol. 12 (2)
Energy Efficiency in Multi-Cell Massive MIMO: Impact of Signal Processing Schemes and Hardware Implementation Parameters
This paper investigates the trade-off between energy efficiency and spectral efficiency in multi-cell massive MIMO systems. The relevance of the study is driven by the need to simultaneously increase throughput and reduce the energy consumption of base stations in next-generation wireless networks, given the growing number of antennas and served users. The purpose of the study is to determine how the number of base-station antennas, the number of served users, linear signal processing schemes, and hardware implementation parameters affect the energy efficiency of multi-cell massive MIMO systems, and to identify configurations that provide the best trade-off between energy and spectral efficiency. Methods. A mathematical model of a massive MIMO system is developed, taking into account the number of base station antennas M, the number of user equipment K, and various linear signal processing schemes, including MR, ZF, RZF, S-MMSE, and M-MMSE. The model incorporates power consumption parameters reflecting hardware implementation characteristics, represented by two different sets of component specifications. The evaluation of energy and spectral efficiency is carried out using simulation-based analysis for various system configurations. Results. The results show that an optimal antenna-to-user ratio of M/K ≈ 3–4 achieves maximum energy efficiency without a significant reduction in spectral efficiency. It is demonstrated that the M-MMSE and S-MMSE algorithms provide the highest energy efficiency performance with moderate computational complexity, particularly when improved hardware components are employed. The obtained results confirm the existence of a pronounced energy efficiency optimum as the number of base station antennas increases . The novelty of this work lies in the comprehensive analysis of the energy–spectral efficiency trade-off in multi-cell massive MIMO systems while jointly accounting for linear signal processing schemes and hardware implementation parameters, which enables the formulation of practical recommendations for base station configuration under technological constraints. Practical significance. The findings of this study can be applied to the design and optimization of energy-efficient multi-cell massive MIMO systems for next-generation wireless communication networks, taking into account hardware implementation constraints and quality-of-service requirements.
- 8 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36 (1) , 350-362
Integrating Disparity Confidence Estimation Into Relative Depth Prior-Guided Unsupervised Stereo Matching
Unsupervised stereo matching has garnered significant attention for its independence from costly disparity annotations. Typical unsupervised methods rely on the multi-view consistency assumption for training networks, which suffer considerably from stereo matching ambiguities, such as repetitive patterns and texture-less regions. A feasible solution lies in transferring 3D geometric knowledge from a relative depth map to the stereo matching networks. However, existing knowledge transfer methods learn depth ranking information from randomly built sparse correspondences, which makes inefficient utilization of 3D geometric knowledge and introduces noise from mistaken disparity estimates. This work proposes a novel unsupervised learning framework to address these challenges, which comprises a plug-and-play disparity confidence estimation algorithm and two depth prior-guided loss functions. Specifically, the local coherence consistency between neighboring disparities and their corresponding relative depths is first checked to obtain disparity confidence. Afterwards, quasi-dense correspondences are built using only confident disparity estimates to facilitate efficient depth ranking learning. Finally, a dual disparity smoothness loss is proposed to boost stereo matching performance at disparity discontinuities. Experimental results demonstrate that our method achieves state-of-the-art stereo matching accuracy on the KITTI Stereo benchmarks among all unsupervised stereo matching methods.
- 8 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36 (1) , 350-362
Integrating Disparity Confidence Estimation Into Relative Depth Prior-Guided Unsupervised Stereo Matching
Unsupervised stereo matching has garnered significant attention for its independence from costly disparity annotations. Typical unsupervised methods rely on the multi-view consistency assumption for training networks, which suffer considerably from stereo matching ambiguities, such as repetitive patterns and texture-less regions. A feasible solution lies in transferring 3D geometric knowledge from a relative depth map to the stereo matching networks. However, existing knowledge transfer methods learn depth ranking information from randomly built sparse correspondences, which makes inefficient utilization of 3D geometric knowledge and introduces noise from mistaken disparity estimates. This work proposes a novel unsupervised learning framework to address these challenges, which comprises a plug-and-play disparity confidence estimation algorithm and two depth prior-guided loss functions. Specifically, the local coherence consistency between neighboring disparities and their corresponding relative depths is first checked to obtain disparity confidence. Afterwards, quasi-dense correspondences are built using only confident disparity estimates to facilitate efficient depth ranking learning. Finally, a dual disparity smoothness loss is proposed to boost stereo matching performance at disparity discontinuities. Experimental results demonstrate that our method achieves state-of-the-art stereo matching accuracy on the KITTI Stereo benchmarks among all unsupervised stereo matching methods.
- Open Access
Integrating Disparity Confidence Estimation into Relative Depth Prior-Guided Unsupervised Stereo Matching
Unsupervised stereo matching has garnered significant attention for its independence from costly disparity annotations. Typical unsupervised methods rely on the multi-view consistency assumption for training networks, which suffer considerably from stereo matching ambiguities, such as repetitive patterns and texture-less regions. A feasible solution lies in transferring 3D geometric knowledge from a relative depth map to the stereo matching networks. However, existing knowledge transfer methods learn depth ranking information from randomly built sparse correspondences, which makes inefficient utilization of 3D geometric knowledge and introduces noise from mistaken disparity estimates. This work proposes a novel unsupervised learning framework to address these challenges, which comprises a plug-and-play disparity confidence estimation algorithm and two depth prior-guided loss functions. Specifically, the local coherence consistency between neighboring disparities and their corresponding relative depths is first checked to obtain disparity confidence. Afterwards, quasi-dense correspondences are built using only confident disparity estimates to facilitate efficient depth ranking learning. Finally, a dual disparity smoothness loss is proposed to boost stereo matching performance at disparity discontinuities. Experimental results demonstrate that our method achieves state-of-the-art stereo matching accuracy on the KITTI Stereo benchmarks among all unsupervised stereo matching methods.
Open AccessIntegrating Disparity Confidence Estimation into Relative Depth Prior-Guided Unsupervised Stereo Matching
Unsupervised stereo matching has garnered significant attention for its independence from costly disparity annotations. Typical unsupervised methods rely on the multi-view consistency assumption for training networks, which suffer considerably from stereo matching ambiguities, such as repetitive patterns and texture-less regions. A feasible solution lies in transferring 3D geometric knowledge from a relative depth map to the stereo matching networks. However, existing knowledge transfer methods learn depth ranking information from randomly built sparse correspondences, which makes inefficient utilization of 3D geometric knowledge and introduces noise from mistaken disparity estimates. This work proposes a novel unsupervised learning framework to address these challenges, which comprises a plug-and-play disparity confidence estimation algorithm and two depth prior-guided loss functions. Specifically, the local coherence consistency between neighboring disparities and their corresponding relative depths is first checked to obtain disparity confidence. Afterwards, quasi-dense correspondences are built using only confident disparity estimates to facilitate efficient depth ranking learning. Finally, a dual disparity smoothness loss is proposed to boost stereo matching performance at disparity discontinuities. Experimental results demonstrate that our method achieves state-of-the-art stereo matching accuracy on the KITTI Stereo benchmarks among all unsupervised stereo matching methods.
- 2 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 2549-2553
DepthMatch: Semi-Supervised RGB-D Scene Parsing Through Depth-Guided Regularization
RGB-D scene parsing methods effectively capture both semantic and geometric features of the environment, demonstrating great potential under challenging conditions such as extreme weather and low lighting. However, existing RGB-D scene parsing methods predominantly rely on supervised training strategies, which require a large amount of manually annotated pixel-level labels that are both time-consuming and costly. To overcome these limitations, we introduce DepthMatch, a semi-supervised learning framework that is specifically designed for RGB-D scene parsing. To make full use of unlabeled data, we propose complementary patch mix-up augmentation to explore the latent relationships between texture and spatial features in RGB-D image pairs. We also design a lightweight spatial prior injector to replace traditional complex fusion modules, improving the efficiency of heterogeneous feature fusion. Furthermore, we introduce depth-guided boundary loss to enhance the model's boundary prediction capabilities. Experimental results demonstrate that DepthMatch exhibits high applicability in both indoor and outdoor scenes, achieving state-of-the-art results on the NYUv2 dataset and ranking first on the KITTI Semantics benchmark. Our source code will be publicly available at mias.group/DepthMatch.
- 2 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Signal Processing Letters
- Vol. 32 (10709908) , 2549-2553
DepthMatch: Semi-Supervised RGB-D Scene Parsing Through Depth-Guided Regularization
RGB-D scene parsing methods effectively capture both semantic and geometric features of the environment, demonstrating great potential under challenging conditions such as extreme weather and low lighting. However, existing RGB-D scene parsing methods predominantly rely on supervised training strategies, which require a large amount of manually annotated pixel-level labels that are both time-consuming and costly. To overcome these limitations, we introduce DepthMatch, a semi-supervised learning framework that is specifically designed for RGB-D scene parsing. To make full use of unlabeled data, we propose complementary patch mix-up augmentation to explore the latent relationships between texture and spatial features in RGB-D image pairs. We also design a lightweight spatial prior injector to replace traditional complex fusion modules, improving the efficiency of heterogeneous feature fusion. Furthermore, we introduce depth-guided boundary loss to enhance the model's boundary prediction capabilities. Experimental results demonstrate that DepthMatch exhibits high applicability in both indoor and outdoor scenes, achieving state-of-the-art results on the NYUv2 dataset and ranking first on the KITTI Semantics benchmark. Our source code will be publicly available at mias.group/DepthMatch.
- Open Access
DepthMatch: Semi-Supervised RGB-D Scene Parsing through Depth-Guided Regularization
RGB-D scene parsing methods effectively capture both semantic and geometric features of the environment, demonstrating great potential under challenging conditions such as extreme weather and low lighting. However, existing RGB-D scene parsing methods predominantly rely on supervised training strategies, which require a large amount of manually annotated pixel-level labels that are both time-consuming and costly. To overcome these limitations, we introduce DepthMatch, a semi-supervised learning framework that is specifically designed for RGB-D scene parsing. To make full use of unlabeled data, we propose complementary patch mix-up augmentation to explore the latent relationships between texture and spatial features in RGB-D image pairs. We also design a lightweight spatial prior injector to replace traditional complex fusion modules, improving the efficiency of heterogeneous feature fusion. Furthermore, we introduce depth-guided boundary loss to enhance the model's boundary prediction capabilities. Experimental results demonstrate that DepthMatch exhibits high applicability in both indoor and outdoor scenes, achieving state-of-the-art results on the NYUv2 dataset and ranking first on the KITTI Semantics benchmark.
Open AccessDepthMatch: Semi-Supervised RGB-D Scene Parsing through Depth-Guided Regularization
RGB-D scene parsing methods effectively capture both semantic and geometric features of the environment, demonstrating great potential under challenging conditions such as extreme weather and low lighting. However, existing RGB-D scene parsing methods predominantly rely on supervised training strategies, which require a large amount of manually annotated pixel-level labels that are both time-consuming and costly. To overcome these limitations, we introduce DepthMatch, a semi-supervised learning framework that is specifically designed for RGB-D scene parsing. To make full use of unlabeled data, we propose complementary patch mix-up augmentation to explore the latent relationships between texture and spatial features in RGB-D image pairs. We also design a lightweight spatial prior injector to replace traditional complex fusion modules, improving the efficiency of heterogeneous feature fusion. Furthermore, we introduce depth-guided boundary loss to enhance the model's boundary prediction capabilities. Experimental results demonstrate that DepthMatch exhibits high applicability in both indoor and outdoor scenes, achieving state-of-the-art results on the NYUv2 dataset and ranking first on the KITTI Semantics benchmark.
- 22 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (6) , 5895-5902
Self-Supervised Diffusion-Based Scene Flow Estimation and Motion Segmentation With 4D Radar
Scene flow estimation (SFE) and motion segmentation (MOS) using 4D radar are emerging yet challenging tasks in robotics and autonomous driving applications. Existing LiDAR- or RGB-D-based point cloud processing methods often deliver suboptimal performance on radar data due to radar signals' highly sparse, noisy, and artifact-prone nature. Furthermore, for radar-based SFE and MOS, the lack of annotated datasets further aggravates these challenges. To address these issues, we propose a novel self-supervised framework that exploits denoising diffusion models to effectively handle radar noise inputs and predict point-wise scene flow and motion status simultaneously. To extract key features from the raw input, we design a transformer-based feature encoder tailored to address the sparsity of 4D radar data. Additionally, we generate self-supervised segmentation signals by exploiting the discrepancy between robust rigid ego-motion estimates and scene flow predictions, thereby eliminating the need for manual annotations. Experimental evaluations on the View-of-Delft (VoD) dataset and TJ4DRadSet demonstrate that our method achieves state-of-the-art performance for both radar-based SFE and MOS.
- 22 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 10 (6) , 5895-5902
Self-Supervised Diffusion-Based Scene Flow Estimation and Motion Segmentation With 4D Radar
Scene flow estimation (SFE) and motion segmentation (MOS) using 4D radar are emerging yet challenging tasks in robotics and autonomous driving applications. Existing LiDAR- or RGB-D-based point cloud processing methods often deliver suboptimal performance on radar data due to radar signals' highly sparse, noisy, and artifact-prone nature. Furthermore, for radar-based SFE and MOS, the lack of annotated datasets further aggravates these challenges. To address these issues, we propose a novel self-supervised framework that exploits denoising diffusion models to effectively handle radar noise inputs and predict point-wise scene flow and motion status simultaneously. To extract key features from the raw input, we design a transformer-based feature encoder tailored to address the sparsity of 4D radar data. Additionally, we generate self-supervised segmentation signals by exploiting the discrepancy between robust rigid ego-motion estimates and scene flow predictions, thereby eliminating the need for manual annotations. Experimental evaluations on the View-of-Delft (VoD) dataset and TJ4DRadSet demonstrate that our method achieves state-of-the-art performance for both radar-based SFE and MOS.
- 26 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-9
SNE-RoadSegV2: Advancing Heterogeneous Feature Fusion and Fallibility Awareness for Freespace Detection
Feature-fusion networks with duplex encoders have proven to be an effective technique to solve the road freespace detection problem. However, despite the compelling results achieved by previous research efforts, the exploration of adequate and discriminative heterogeneous feature fusion, as well as the development of fallibility-aware loss functions, remains relatively scarce. This article makes several significant contributions to address these limitations: 1) it presents a novel heterogeneous feature fusion block, comprising a holistic attention module, a heterogeneous feature contrast descriptor, and an affinity-weighted feature recalibrator, enabling a more in-depth exploitation of the inherent characteristics of the extracted features, 2) it incorporates both inter-scale and intra-scale skip connections into the decoder architecture, while eliminating redundant ones, leading to both improved accuracy and computational efficiency, and 3) it introduces two fallibility-aware loss functions that separately focus on semantic-transition and depth-inconsistent regions, collectively contributing to greater supervision during model training. Our proposed SNE-RoadSegV2, which incorporates all these innovative components, demonstrates superior performance in comparison to all other freespace detection algorithms across multiple public datasets.
- 26 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-9
SNE-RoadSegV2: Advancing Heterogeneous Feature Fusion and Fallibility Awareness for Freespace Detection
Feature-fusion networks with duplex encoders have proven to be an effective technique to solve the road freespace detection problem. However, despite the compelling results achieved by previous research efforts, the exploration of adequate and discriminative heterogeneous feature fusion, as well as the development of fallibility-aware loss functions, remains relatively scarce. This article makes several significant contributions to address these limitations: 1) it presents a novel heterogeneous feature fusion block, comprising a holistic attention module, a heterogeneous feature contrast descriptor, and an affinity-weighted feature recalibrator, enabling a more in-depth exploitation of the inherent characteristics of the extracted features, 2) it incorporates both inter-scale and intra-scale skip connections into the decoder architecture, while eliminating redundant ones, leading to both improved accuracy and computational efficiency, and 3) it introduces two fallibility-aware loss functions that separately focus on semantic-transition and depth-inconsistent regions, collectively contributing to greater supervision during model training. Our proposed SNE-RoadSegV2, which incorporates all these innovative components, demonstrates superior performance in comparison to all other freespace detection algorithms across multiple public datasets.
- Open Access
- 1 December 2023
- Published by Kotelnikov Institute of Radioengineering and Electronics of Russian Academy of Sciences in Journal of Radio Electronics
- Vol. 2023 (12)
Comparative analysis of accuracy of frequency synchronization algorithms for satellite communication system based on DVB-S2 signal
The paper investigates the schemes of frequency synchronization of signal reception of satellite communication system based on DVB-S2 standard with different combinations of coarse and fine frequency synchronization algorithms, namely, the following pairs of algorithms are considered: «Edge band» and «Delay & Multiply», «Delay & Multiply» and «L&R», «Edge band» and «L&R», where the first algorithm in each pair performs the task of coarse and the second algorithm performs the task of fine frequency synchronization. The accuracy of the investigated frequency synchronization schemes is evaluated using the Root Mean Square Error (RMSE) metric of frequency offset estimation in the channel. In addition, one of the performance metrics of the algorithm is the estimation of synchronization time. The frequency shift in the signal transmission channel is set constant and equal to 20% of the signal bandwidth. Using combinations of Non-Data-Aided (blind) algorithms for coarse frequency shift estimation and Data-Aided algorithms for fine frequency synchronization, the synchronization time can be reduced by almost a factor of 2, while maintaining high accuracy of RMS estimation.
- 1 December 2023
- Published by Kotelnikov Institute of Radioengineering and Electronics of Russian Academy of Sciences in Journal of Radio Electronics
- Vol. 2023 (12)
Open AccessComparative analysis of accuracy of frequency synchronization algorithms for satellite communication system based on DVB-S2 signal
The paper investigates the schemes of frequency synchronization of signal reception of satellite communication system based on DVB-S2 standard with different combinations of coarse and fine frequency synchronization algorithms, namely, the following pairs of algorithms are considered: «Edge band» and «Delay & Multiply», «Delay & Multiply» and «L&R», «Edge band» and «L&R», where the first algorithm in each pair performs the task of coarse and the second algorithm performs the task of fine frequency synchronization. The accuracy of the investigated frequency synchronization schemes is evaluated using the Root Mean Square Error (RMSE) metric of frequency offset estimation in the channel. In addition, one of the performance metrics of the algorithm is the estimation of synchronization time. The frequency shift in the signal transmission channel is set constant and equal to 20% of the signal bandwidth. Using combinations of Non-Data-Aided (blind) algorithms for coarse frequency shift estimation and Data-Aided algorithms for fine frequency synchronization, the synchronization time can be reduced by almost a factor of 2, while maintaining high accuracy of RMS estimation.
- 1 January 2022
- Published by Springer Nature
- p. 338-348
Transient Behaviour of Finite-Source Single-Line Queueing Systems with Jumps of Network Traffic
- 1 January 2022
- Published by Springer Nature
- p. 338-348
Transient Behaviour of Finite-Source Single-Line Queueing Systems with Jumps of Network Traffic
- 24 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Low Complexity DVB-S2X Frequency Synchronization for LEO Satellites
This paper addresses the problem of frequency synchronization for LEO satellite systems based on DVB-S2/S2X standard. This problem is rather crucial nowadays, since LEO satellite systems gain more and more popularity (e.g. Starlink) and at Low Earth Orbit the Doppler shift can change at a relatively high speed (tens of kHz per second). We demonstrate that the simple geostationary satellite synchronization algorithms, such as second-order PLL and the well-known Fitz algorithm, fail to solve this problem. We give recommendations on an efficient frequency synchronization systems design, which could solve the problem and keep the computational complexity of the entire scheme low at the same time.
- 24 November 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Low Complexity DVB-S2X Frequency Synchronization for LEO Satellites
This paper addresses the problem of frequency synchronization for LEO satellite systems based on DVB-S2/S2X standard. This problem is rather crucial nowadays, since LEO satellite systems gain more and more popularity (e.g. Starlink) and at Low Earth Orbit the Doppler shift can change at a relatively high speed (tens of kHz per second). We demonstrate that the simple geostationary satellite synchronization algorithms, such as second-order PLL and the well-known Fitz algorithm, fail to solve this problem. We give recommendations on an efficient frequency synchronization systems design, which could solve the problem and keep the computational complexity of the entire scheme low at the same time.