检索结果-内蒙古大学图书馆

Efficient Hierarchical Feature Collaboration Transformer for Image Inpainting

IEEE Transactions on Multimedia 2025年

作者： Zhang, Dengyong Fu, Nuo Liao, Xin Chen, Jiaxin Yang, Hengfu Yang, Gaobo Changsha University of Science and Technology School of Computer and Communication Engineering Changsha410004 China Hunan University College of Computer Science and Electronic Engineering Changsha410082 China Hunan First Normal University School of Computer Science Changsha410205 China

Existing image inpainting methods face limitations in detail restoration. Although transformer-based models have made certain progress recently, the lack of hierarchical feature interaction and insufficient consideration of the importance of features at different network levels lead to semantic ambiguity in image reconstruction. To enhance the visual quality and accuracy of image inpainting, we adopt a multi-level feature fusion approach and propose a novel, efficient hierarchical feature collaboration transformer (HFCT). Our approach comprises two modules: dual stream gated feature fusion (DSGF) and region-separated attention module (RSAM), effectively capturing features at different levels of the network and enhancing inter-level information exchange. The DSGF module uses soft gating to fuse primary and advanced features, strengthening the connection from local to global consistency and reducing artifacts. The RSAM module resolves attention isolation issues in feature fusion through region-separated attention, strengthening the understanding of feature relationships, capturing more image semantics, and improving restoration accuracy. Extensive experiments on the Paris StreetView, CelebA-HQ, and Places2 benchmark datasets demonstrate that our proposed method achieves superior image inpainting quality compared to several state-of-the-art inpainting algorithms. Please refer to the project page: https://***/csfunuo/HFCT. © 2024 IEEE.

关键词： Restoration

来源：评论

学校读者我要写书评

暂无评论

Driver Drowsiness Detection Using Swin Transformer and Diffusion Models for Robust Image Denoising

引用

IEEE Access 2025年 13卷 71880-71907页

作者： El-Nabi, Samy Abd Ibrahim, Ahmed F. El-Rabaie, El-Sayed M. Hassan, Osama F. Soliman, Naglaa F. Ramadan, Khalil F. El-Shafai, Walid Faculty of Computer Science and Engineering Department of Artificial Intelligence Engineering South Sinai 46511 Egypt Menoufia University Faculty of Electronic Engineering Department of Electronics and Electrical Communications Engineering Menouf32952 Egypt Suez Canal University Faculty of Computers and Informatics Information Systems Department Ismailia41522 Egypt Princess Nourah bint Abdulrahman University College of Computer and Information Sciences Department of Information Technology P.O. Box 84428 Riyadh11671 Saudi Arabia Computer Science Department Riyadh11586 Saudi Arabia

With the rapid development of intelligent transportation systems and growing emphasis on driver safety, real-time detection of driver drowsiness has become a critical area of research. This study presents a robust and scalable driver drowsiness detection framework that integrates a Swin Transformer-based deep learning model with a diffusion model for image denoising. While conventional convolutional neural networks (CNNs) are effective in standard vision tasks, they often suffer performance degradation in real-world driving scenarios due to noise, poor lighting, motion blur, and adversarial attacks. To address these challenges, the proposed model focuses on eye-state detection, specifically, prolonged eye closure, as a primary indicator of driver disengagement and fatigue. Our system introduces a novel preprocessing stage using a denoising diffusion model built on a U-Net encoder-decoder architecture, effectively mitigating the impact of Gaussian noise and adversarial perturbations. Additionally, we incorporate adversarial training with Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) attacks, demonstrating significant improvements in classification accuracy and resilience. Evaluations are conducted on two benchmark datasets, Eye-Blink and Closed Eyes in the Wild (CEW), under both clean and noisy conditions. Comparative experiments show that the proposed system outperforms several state-of-the-art models, including ViT, ResNet50V2, InceptionV3, MobileNet, DenseNet169, and VGG19, in terms of accuracy (up to 99.82%), PSNR (up to 41.61 dB), and SSIM (up to 0.984), while maintaining competitive inference times suitable for practical deployment. Moreover, a detailed sensitivity analysis of data augmentation strategies reveals that techniques such as rotation and horizontal flip substantially enhance the model’s generalization across variable visual inputs. The system also demonstrates improved robustness under real-world black-box scenarios and adver

关键词： Sensitivity analysis

来源：评论

学校读者我要写书评

暂无评论

Efficient Intrusion Detection in AMI Systems Based on Federated Semi-supervised Learning

引用

IEEE Transactions on Network science and engineering 2025年

作者： Xia, Zhuoqun Tang, Haidong Hu, Zhenzhen Zhou, Hongmei Changsha University of Science and Technology School of Computer and Communication Engineering China University of South China College of Computer Science China

The advanced metering infrastructure (AMI) network is the most critical part of a smart grid, and it faces serious security challenges from network attacks. Federated learning (FL) is a common method for addressing network security problems;however, it has several shortcomings, such as difficulty in obtaining labeled data and high communication costs. Therefore, in this paper, an efficient intrusion detection method based on federated semi-supervised learning is proposed to improve the efficiency of AMI network intrusion detection and reduce communication overhead. First, an intrusion detection framework based on federated distillation (FD) is established to address intense assumption dependency problems and reduce communication overhead. Then, an efficient intrusion detection algorithm is used to improve the classification performance. Finally, the designed deep convolutional generative adversarial network (DCGAN) model is used to obtain high-quality sample data. The experimental results show that the scheme achieves an accuracy of 99.58%, a communication overhead of 75 MB, and a reduction in the false-positive rate of 6%. © 2013 IEEE.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Feature Selection in Socio-Economic Analysis: A Multi-Method Approach for Accurate Predictive Outcomes

引用

International Journal of Crowd science 2025年第1期9卷 64-78页

作者： Al-Qerem, Ahmad Ali, Ali Mohd Jebreen, Issam Nabot, Ahmad Rajab, Mohammed Alauthman, Mohammad Aldweesh, Amjad Aburub, Faisal Alangari, Someah Alzgol, Musab Computer Science Department Faculty of Information Technology Zarqa University Zarqa13110 Jordan Communications and Computer Engineering Department Faculty of Engineering Al-Ahliyya Amman University Amman19328 Jordan University Headquarter University of Anbar Ramadi31001 Iraq Department of Information Security University of Petra Amman11196 Jordan College of Computing and Information Technology Shaqra University Shaqra11911 Saudi Arabia Department of Business Intelligence and Data Analytics University of Petra Amman11196 Jordan Computer Information Systems Department Faculty of Information Technology Isra University Amman11622 Jordan

Feature selection is a cornerstone in advancing the accuracy and efficiency of predictive models, particularly in nuanced domains like socio-economic analysis. This study explores nine distinct feature selection methods, utilizing a heart disease dataset as a representative model for complex socio-economic systems. Our findings identified four universally recognized features as critical across all selection methods. However, the divergence in significance attributed to other features by different methods underscores the inherent variability in selection techniques. When the top four features were incorporated into twelve classification models, a noticeable surge in predictive accuracy was observed, emphasizing their foundational role in enhancing model outcomes. The variations among methods stress the need for a methodical and discerning approach to feature selection, especially in data-rich socio-economic landscapes. As we venture further into an era defined by data-driven decision-making, rigour and precision in feature selection become indispensable. Future research should extend this approach to broader datasets, ensuring the robustness and adaptability of our findings. © The author(s) 2025.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Wideband Circularly-Polarized Hexagonal SIW Cavity-Backed Slot Antenna With Enhanced Bandwidth and Compact Design

引用

IEEE Access 2025年 13卷 80407-80415页

作者： Sarani, Vahid Mousa Ali, Esraa Shariatifar, Milad Virdee, Bal S. Mariyanayagam, Dion Rashid, Nasr Alibakhshikenari, Mohammad Dalarsson, Mariana Ferdowsi University of Mashhad Electrical Engineering Department Mashhad*** Iran Al-Ahliyya Amman University Communications and Computer Engineering Department Amman19111 Jordan University of Rome "Tor Vergata " Electronics Engineering Department Rome00133 Italy London Metropolitan University Center for Communications Technology LondonN7 8DB United Kingdom Jouf University College of Engineering Electrical Engineering Department Al Jowf Sakaka72388 Saudi Arabia Al-Azhar University Faculty of Engineering Department of Electrical Engineering Nasr City Cairo11884 Egypt Dogus University Department of Electrical and Electronics Engineering Istanbul Umraniye34775 Turkey KTH Royal Institute of Technology School of Electrical Engineering and Computer Science Stockholm100 44 Sweden

This paper presents a novel wideband circularly polarized (CP) cavity-backed slot antenna based on Substrate Integrated Waveguide (SIW) technology, designed for compact and high-efficiency performance. The proposed antenna utilizes a hexagonal SIW cavity to simultaneously excite two closely spaced resonant modes (TM110 and TM210), resulting in enhanced bandwidth for linear polarization (LP). To achieve circular polarization, a passive, single-layer linear-to-circular polarization converter is integrated above the cavity, offering a structurally simple and PCB-compatible solution. Unlike conventional CP designs that rely on complex feeding networks or multilayered structures, this configuration maintains a planar profile and efficient performance. A fabricated prototype demonstrates strong agreement between simulation and measurement, achieving a peak gain of 9.2 dBic and a 14% axial ratio (AR) bandwidth. These results highlight the antenna’s suitability for modern wireless systems requiring wideband CP functionality, including satellite communications, 5G modules, and compact embedded devices. © 2013 IEEE.

关键词： Circular polarization

来源：评论

学校读者我要写书评

暂无评论

Radio resource management and path planning in intelligent transportation systems via reinforcement learning for environmental sustainability

Radio resource management and path planning in intelligent t...

引用

International Conference on Innovation and Technological Advances for Sustainability, ITAS 2023

作者： Norouzi, S. Azarasa, N. Abedi, M.R. Mokari, N. Seyedabrishami, S.E. Saeedi, H. Jorswieck, E.A. Department of Electrical and Computer Engineering Tarbiat Modares University Tehran Iran Department of Civil and Environmental Engineering Tarbiat Modares University Tehran Iran College of Engineering Doha University for Science and Technology Doha Qatar Institute of Communications Technology TU Braunschweig Lower Saxony Germany

ISBN: (纸本)9781032803722

Efficient and dynamic path planning has become an important topic for urban areas with larger density of connected vehicles (CV) which results in reduction of travel time and directly contributes to environmental sustainability through reducing energy consumption. CVs exploit the cellular wireless vehicle-to-everything (C-V2X) communication technology to disseminate the vehicle-to-infrastructure (V2I) messages to the Base-station (BS) to improve situation awareness on urban roads. In this paper, we investigate radio resource management (RRM) in such a framework to minimize the age of information (AoI) so as to enhance path planning results. We use the fact that V2I messages with lower AoI value result in less error in estimating the road capacity and more accurate path planning. Through simulations, we compare road travel times and volume over capacity (V/C) against different levels of AoI and demonstrate the promising performance of the proposed framework. © 2025 The Author(s).

关键词： Resource allocation

来源：评论

学校读者我要写书评

暂无评论

EBCFL: An Efficient BlockChain-based Federated Learning Framework for Internet of Vehicles

引用

IEEE Network 2025年

作者： Cao, Dun Liao, Yonghui Alqahtani, Fayez Wang, Jin School of Computer and Communication Engineering Changsha University of Science and Technology China Changsha University of Science and Technology China Software Engineering Department College of Computer and Information Sciences King Saud University Saudi Arabia Hunan University of Science and Technology China

Federated Learning (FL) can train a shared model while protecting vehicle data privacy, which has significant advantages in improving driving safety and efficiency. However, there are still many challenges in applying FL to Internet of Vehicles (IoV). First, the varying computing capabilities of moving vehicles result in slow convergence speed due to underperforming vehicles. Second, malicious vehicles may tamper with model parameters before uploading them to the central server, which is itself vulnerable to single-point failure. To solve the above challenges, we design an efficient and decentralized blockchain FL framework (Efficient BlockChain-based Federated Learning Framework, EBCFL). This framework effectively prevents possible malicious behaviors in the vehicular network nodes and improves the efficiency of FL through a well-designed security verification mechanism. The experimental results show that EBCFL outperforms other schemes in terms of average test accuracy and average total time cost, indicating that EBCFL significantly improves the learning efficiency while bearing an acceptable additional time cost for blockchain communication, demonstrating its strong ability to guard against malicious behaviours and ensure robustness. © 1986-2012 IEEE.

关键词： Blockchain Federated learning IoV Security verification

来源：评论

学校读者我要写书评

暂无评论

A Contend Node-Restricted Receiver-Initiated MAC Protocol With Consecutive Packet Transmission

IEEE Transactions on Green Communications and Networking

引用

IEEE Transactions on Green communications and Networking 2025年

作者： Liu, Yuxin Ota, Kaoru Dong, Mianxiong Wang, Ruoyu Changsha University College of Computer Science and Engineering Changsha410022 China Muroran Institute of Technology Department of Sciences and Informatics Japan Central South University School of Computer Science and Engineering Changsha410083 China

The medium access control (MAC) protocol plays a crucial role in wireless communications, influencing key performance metrics, such as energy consumption, delay, collision, and throughput. This paper introduces the Contend Node Restricted Joint Consecutive Packet Transmissions Receiver-Initiated (CNR-CPT-RI) protocol, specifically designed to minimize collisions and reduce energy consumption in wake-up radio-enabled wireless sensor networks (WSNs). The CNR-CPT-RI protocol is an effective communication protocol that is mainly applied with nodes in the one-hop range of the sink. Since the sink has the ability to control each node, it becomes feasible to manage the nodes in a way that effectively reduces collisions and enhances network lifetime as well as throughput. Moreover, the CNR-CPT-RI protocol can adaptively adjust the number of nodes participating in channel contention, maintaining it within an optimal range to balance reduced collisions, decreased energy consumption, and sustained high throughput. Furthermore, the winning node in channel contention employs Consecutive Packet Transmissions (CPTs) method to further diminish competition. Comprehensive theoretical and experimental results demonstrate that the number of collisions of the CNR-CPT-RI protocol is only 42.4%, 35.9%, and 32.5% of the baseline protocol when the number of nodes is 10, 20, and 30, respectively. The energy consumption of the CNR-CPT-RI protocol is reduced by 5.7%, 15%, and 22.6%, and the delay is decreased by 6.7%, 16%, and 32%, respectively, compared with that of the existing protocol. © 2017 IEEE.

关键词： Internet protocols

来源：评论

学校读者我要写书评

暂无评论

Semi-Blind Channel Estimation and Achievable Rate Analysis for Uplink RIS-Enhanced Multi-User Networks

IEEE Open Journal of Vehicular Technology

引用

IEEE Open Journal of Vehicular technology 2025年 6卷 1108-1139页

作者： Farghaly, Samar I. Ismail, Mohamed I. Fouda, Mostafa M. Alwakeel, Ahmed S. Tanta University Department of Electronics and Electrical Communications Engineering Faculty of Engineering Tanta31111 Egypt Idaho State University Department of Electrical and Computer Engineering College of Science and Engineering PocatelloID83209 United States Modern Academy for Engineering Department of Electronic Engineering and Communication Technology Cairo Governorate 4411602 Egypt Center for Advanced Energy Studies Idaho FallsID83401 United States Suez University Department of Electrical Engineering Faculty of Engineering Suez43221 Egypt

Future wireless networks could benefit from the energy-efficient, low-latency, and scalable deployments that Reconfigurable Intelligent Surfaces (RISs) offer. However, the creation of an effective low overhead channel estimate technique is a major obstacle in RIS-assisted systems, especially given the high number of RIS components and intrinsic hardware constraints. This research examines the uplink of a RIS-empowered multi-user MIMO communication system and presents a novel semi-blind channel estimate approach. Unlike current approaches, which rely on pilot-based channel estimation, our methodology uses data to estimate channels, considerably enhancing the achievable rate. We provide a closed-form deterministic expression for the uplink achievable rate in actual settings where the channel state information (CSI) must be estimated rather than assumed perfect. The results of the simulations show that the formula obtained is accurate, with a close alignment between the deterministic and actual achievable rates (generally between 2 5% deviations). The proposed approach outperforms traditional approaches, resulting in rate increases of up to 35–40%, especially in instances with more RIS elements. These findings illustrate RIS technology's tremendous potential to improve system capacity and coverage, providing useful insights for optimizing RIS adoption in future wireless networks. © 2020 IEEE.

关键词： Channel state information

来源：评论

学校读者我要写书评

暂无评论

Towards Privacy-Enhanced and Robust Clustered Federated Learning

引用

IEEE Transactions on Mobile Computing 2025年

作者： Xu, Yang Tan, Yunlin Zhang, Cheng Sun, Peng Zhang, Yibang Ren, Ju Jiang, Hongbo Zhang, Yaoxue Hunan University College of Cyber Science and Technology Changsha410082 China Tsinghua University Department of Computer Science and Technology BNRist Beijing100084 China Hunan University College of Computer Science and Electronic Engineering Changsha410082 China

Clustered federated learning (CFL) leverages data distribution similarities to cluster clients, facilitating personalized model training under data heterogeneity. However, most existing CFL schemes pose potential privacy risks for clients (e.g., gradient inversion attacks) as they rely on individual gradients for clustering. This also renders them incompatible with secure aggregation mechanisms that are widely employed in federated learning for privacy protection. Moreover, CFL introduces the risk of malicious clients dominating several clusters and conducting poisoning attacks therein, thereby threatening secure model training. To address these issues, we propose ProCFL, a Privacy-Enhanced and Robust CFL framework incorporating gradient-free clustering and peer validation. Specifically, we first design a new protocol for measuring data distribution similarity among clients without using their gradient information. Then, we transform the client clustering process into a weighted set covering problem and introduce a diversity-optimized clustering algorithm to achieve near-optimal clustering results while eliminating any need for prior knowledge. Furthermore, we develop a post-hoc detection mechanism that employs peer validation to identify and discard malicious client models. Extensive experimental evaluation of ProCFL validates its superior model robustness and accuracy performance compared to existing schemes. © 2025 IEEE. All rights reserved.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：