最新刊期
卷 42 ， 期 2 ， 2026
- 摘要：With the advancement of the space-integrated-ground network, device-to-satellite communication is transitioning from concept to reality. Achieving stable and clear voice communication with limited satellite link resources is a key challenge for the industry. Due to the bandwidth limitations, high path loss, and high transmission delays of satellite channels, speech codecs used in terrestrial networks are not directly adaptable to satellite communication scenarios. Therefore, low bit-rate speech codec is crucial for satellite voice services. Based on this, the low bit-rate speech codec technologies for satellite communication were systematically summarized, the principles, characteristics, and performance evaluations of mainstream technical routes were introduced, the advantages and disadvantages of each method were analyzed, and future research directions were prospected.关键词：satellite communication;low bit-rate speech codec;speech quality6060|5891|0更新时间：2026-03-05
Expert Views
- 摘要：To address the semantic redundancy caused by nonlinear inter-frame content changes during video temporal resampling, an adaptive resampling method based on self-supervised feature embedding and clustering was proposed. In this method, features from video frames were extracted using a pre-trained ResNet-18, which was subsequently fine-tuned. Self-supervised metric learning was employed to construct an inter-frame similarity measure, and cosine similarity was used to gauge the resemblance between adjacent frames. A loss function was designed to ensure a smooth manifold distribution for frames from the same video sequence in the embedding space, while similarity between frames from different videos was simultaneously suppressed. Subsequently, the embedded frame features were subjected to temporal data clustering based on manifold bisection points, and the integrity of the video’s first and last frames was ensured. The sampled video sequence was then encoded using H.266/VVC, and the original frames were reconstructed at the decoding end by a frame interpolation network. Through experiments, it was demonstrated that average improvements of approximately 2.3% and 19.4% were achieved on the BDmAP and Pareto mAP metrics, respectively. Furthermore, the computational overhead was found to meet the demands of real-time processing. The proposed approach effectively balances compression efficiency, visual task accuracy, and zero-shot compatibility. It offers a novel solution for video transmission in machine vision scenarios.关键词：machine vision;temporal resampling;video feature embedding;self-supervised clustering;video compression5994|4741|0更新时间：2026-03-05
- 摘要：With the continuous expansion of the utilized electromagnetic spectrum, the wideband sensing capability of reconnaissance receivers is increasingly challenged. The Nyquist folding receiver (NYFR), characterized by its ultra-wideband sensing potential, depends on the estimation of the Nyquist zone (NZ) index to reconstruct signal. However, existing NZ index estimation algorithms fail to achieve satisfactory generalization performance for mixed inputs comprising multiple types of non-cooperative signals. Based on the convolutional blind denoising network (CBDNet) and YOLOv5, a novel parameter estimation algorithm for NYFR output signals was proposed to solve these problems. Firstly, the output signals of the NYFR were transformed into time-frequency representations. Then, CBDNet was used to reconstruct the time-frequency features of signal, and YOLOv5 was used to estimate the corresponding NZ index. Based on the estimation results of NZ index, each signal spectrum was reconstructed, and the unknown carrier frequencies of the sensed signals were obtained. Simulation results validate the effectiveness of the algorithm approach, demonstrating its capability to sense spectrum accurately and estimate carrier frequency for a variety of modulation types from 0～20 GHz, under a sampling rate of 2 GHz. Furthermore, the method exhibits robust performance under spectrum aliasing conditions, enhancing the generalization and adaptability of the NYFR for processing non-cooperative modulated signals.关键词：spectrum sensing;NYFR;NZ index;YOLOv5143|59|0更新时间：2026-03-05
- 摘要：Reconfigurable intelligent surface (RIS) has gained significant attention in millimeter-wave communications due to its advantages, including low power consumption, easy tunability, and enhanced auxiliary communication capabilities. Most existing transmission schemes employ channel state information to design precoding and passive beamforming matrices for RIS. However, this approach incurs substantial pilot overhead, thereby reducing spectral efficiency. To address this issue, a beamforming design scheme for RIS-assisted multiple-input multiple-output (MIMO) systems based on the multi-armed bandit (MAB) algorithm was proposed. The channel covariance matrix was estimated using historical data via the MAB framework, which helped to reduce pilot overhead. Specifically, the passive beamforming matrix design was formulated as an MAB problem and solved using the linear upper confidence bound (LinUCB) algorithm to estimate the channel covariance matrix. The effective spectral efficiency was defined as the reward, while the RIS phase shift vector constitutes the action. The phase shift vector that maximizes the sum of effective spectral efficiency was selected through a hierarchical greedy search algorithm. Simulation results demonstrate that the proposed algorithm effectively reduces pilot overhead and enhances spectral efficiency, thereby confirming its superiority.关键词：reconfigurable intelligent surface;multiple-input multiple-output;multi-armed bandit;LinUCB algorithm98|63|0更新时间：2026-03-05
- 摘要：To address the issues of high link overhead and numerical instability in hybrid precoding design for cell-free massive multiple-input multiple-output (CF-mMIMO) systems operating in the millimeter-wave band, a column-orthogonality-constrained orthogonal matching pursuit (OMP) algorithm adapted to distributed architectures was proposed. In the analog precoding phase, a least squares estimation of signal parameters via rotational invariance technique (LS-ESPRIT) algorithm was employed to directly estimate the angle of arrival from received signals and reconstruct the beamforming codebook, thereby significantly reducing the feedback overhead. In the digital precoding phase, a column orthogonality constraint was introduced, and the optimal scaling factor was derived to reformulate the objective function. This avoided the inversion of ill-conditioned matrices and enhanced both the numerical stability and matching accuracy of the precoding matrix. Simulation results demonstrate that the proposed algorithm consistently achieves stable and significant improvements in spectral efficiency across various access point distributions and system parameter configurations, verifying its robustness and effectiveness in CF-mMIMO systems.关键词：CF-mMIMO;millimeter wave;OMP algorithm;spectral efficiency;hybrid precoding84|51|0更新时间：2026-03-05
- 摘要：In millimeter-wave massive multiple-input multiple-output (MIMO) systems, adaptive fully connected architectures suffered from binary constraints, constant modulus constraints, and insufficient utilization of channel information, resulting in limited spectral and energy efficiency. To address this issue, a hybrid precoding method integrating Gumbel-Softmax and convolutional neural network (CNN) was proposed. Two CNN subnetworks, TsNet and TpsNet, were designed to optimize the switch precoding matrix and the phase shift precoding matrix, respectively. In TsNet, the Gumbel-Softmax method was innovatively introduced to embed discrete binary constraints. TpsNet used a phase layer to constrain the output to the effective phase range of the phase shifter and utilized the C2 layer to satisfy the constant modulus constraint. TsNet and TpsNet were combined in parallel to form a joint network, PCNet, which extracted millimeter-wave channel features using a residual network. The two subnetworks were trained in parallel, sharing residual network parameters to enhance feature consistency, resulting in a near-optimal precoding matrix. Simulation results show that PCNet achieves improved spectral and energy efficiency compared to other competing algorithms. This method significantly enhances system spectral and energy efficiency.关键词：hybrid precoding;adaptive fully-connected architecture;convolutional neural network;residual network;Gumbel-Softmax72|53|0更新时间：2026-03-05
- 摘要：With the rapid development of intelligent connected vehicle (ICV) technology, the cooperative intelligent transportation application standard system has been facing compatibility challenges due to the coexistence of multiple protocol versions. The study focused on the vehicular communication application system, systematically verified and analyzed the inheritance and extension relationships between application scenarios and message sets across different versions, and proposed a compatibility evaluation model based on ASN.1 syntax rules. The study reveals that the Phase 1 message sets exhibit backward compatibility at the syntactic level. However, during the functional iteration of Phase 2 standards, compatibility limitations emerged in some message bodies due to scenario refinement requirements. Based on these findings, the study proposed suggestions for source code management mechanisms and industry deployment strategies, to provide a theoretical foundation for industrial- scale applications.关键词：C-V2X;C-ITS;V2X application;message compatibility;ASN.1 extension47|130|0更新时间：2026-03-05
- 摘要：To address the heterogeneous cryptographic communication problem in the Internet of vehicles (IoV), an online/offline heterogeneous signcryption scheme supporting multiple ciphertext equality tests was proposed to achieve secure communication from a certificateless cryptosystem to a public key infrastructure. The scheme was constructed based on elliptic curve cryptosystem (ECC), and the online/offline signcryption mechanism was employed to reduce the computational overhead on vehicles. A multiple ciphertext equality test was performed in the cloud so that the receiver only needed to download the duplicate ciphertext once, thereby reducing the receiver’s burden. In terms of security, the scheme was proven under the random oracle model (ROM) to satisfy non-repudiation and confidentiality security. Furthermore, the scheme was formally verified using ProVerif and Scyther tools. The ProVerif results show that message confidentiality, identity anonymity, and signature correctness are guaranteed, while the Scyther results show that no effective attack paths are found. Performance analysis demonstrates that the scheme achieves lower computational and communication overhead compared with the existing schemes, and the advantages become more significant with an increasing number of ciphertexts, making it suitable for IoV environments.关键词：IoV;ECC;heterogeneous signcryption;online/offline;multi-ciphertext equality test73|96|0更新时间：2026-03-05
- 摘要：Aiming at the phenomena of insufficient fine-grained texture synthesis, structural repair faults, and semantic detuning, which are commonly found in face image restoration tasks in complex contexts, a face image restoration network based on the fusion enhancement of a dynamic gating mechanism with a self-attention module was proposed. The algorithm captured local details and long-range contextual information by constructing a multilevel dilated convolutional group, and introduced a dual innovative mechanism: (1) the deep dynamic gating mechanism adopted multilayer convolution with batch normalization to achieve spatially adaptive feature selection, replacing the fixed fusion of the traditional residual connection, which significantly enhanced the flexibility and accuracy of feature expression; (2) the self-attention mechanism explicitly modeled global pixel dependencies, which effectively solved the difficulties of structural coherence and fine-grained texture synthesis in large-scale defect repair. Experiments show that, compared with the better comparison algorithm SCAT, this new method improves PSNR and SSIM metrics by an average of 0.382 dB and 0.004 1, and improves FID by an average of 7.81% on three face datasets, namely, FFHQ, CelebA-HQ, and LFW, especially in the scene of large-area occlusion (>50%), the FID decreased by an average of 2.153 4, significantly improving the accuracy of face images in complex backgrounds. It improves the quality of face image restoration under complex backgrounds, especially in generating realistic textures and structural consistency, showing outstanding advantages.关键词：contextual modeling;dynamic feature fusion;dynamic gating mechanism;self-attention mechanism45|202|0更新时间：2026-03-05
- 摘要：To address issues such as feature loss, high computational complexity, and insufficient real-time performance in distributed denial of service (DDoS) attack detection within software-defined networks (SDN), a systematic detection framework was proposed. Firstly, traffic characterization method integrateing dual-granularity information at both flow-level and packet-level was introduced to extract key features of various attack behaviors at multiple scales, thereby enhancing the completeness of traffic representation. Then, a lightweight detection model named DDoSMamba, based on the Mamba architecture, was constructed. By leveraging state space modeling and global receptive field mechanisms, the model reduced computational and memory overhead during sequence modeling. A bidirectional information interaction mechanism was introduced to enhance contextual modeling, while low-rank approximation and subspace feature decomposition strategies were employed to significantly compress parameter size and inference cost. Finally, a two-stage DDoS attack detection method was designed. In the first stage, Tsallis entropy was used to perform rapid filtering based on coarse-grained features, effectively eliminating a large amount of benign traffic. In the second stage, fine-grained features were used for high-precision classification, achieving a balance between fast response and accurate detection. Experiments conducted on the CIC-IDS2019 dataset demonstrate that the proposed method achieves 99.96% and 99.93% detection accuracy for binary and multi-class classification tasks, respectively, with an average inference latency of only 0.067 2 ms and a model size as low as 4.553 8 KB.关键词：SDN;DDoS attack detection;traffic representation;two-stage detection and classification80|134|0更新时间：2026-03-05
- 摘要：To address the issue of insufficient learning caused by the lack of positive samples for large aspect ratio targets in remote sensing images, a shape-adaptive label assignment network for oriented object detection (SALANet) was proposed. Firstly, an aspect ratio sensitivity coefficient was introduced to establish a dynamic mapping relationship between target geometric features and the number of positive samples, alleviating the sample distribution imbalance caused by fixed allocation rules in traditional methods. Secondly, an adaptive label assignment strategy was designed to prioritize high-quality positive samples through intersection over union (IoU) ranking. Finally, a central axis prior was proposed, extending the circular central prior region to a rectangular region along the target’s central axis, thereby enhancing the geometric feature representation capability for large aspect ratio targets. Comparative experiments on the DOTAv1.0 and HRSC2016 datasets demonstrated that SALANet achieved mAP scores of 0.777 1 and 0.932 3, respectively, representing improvements of 8.15% and 2.87% over the baseline method RoI Transformer.关键词：remote sensing image;oriented object detection;large aspect ratio target;shape-adaptive label assignment;central axis prior40|186|0更新时间：2026-03-05
- 摘要：Traditional OFDM frame synchronization methods, such as preamble-sequence-based, pilot-based, and deep learning-based frame synchronization algorithms, are limited to single-channel computing ideas, resource consumption, and high clock requirements in the face of Starlink’s high-speed and high-dynamic communication environment.However, although the multi-channel parallel frame synchronization algorithm based on SDPSK modulation introduces frequency offset correction and local preamble sequence correlation to improve the performance and reduce the clock requirements, it increases the amount of computation and hardware resources. In view of the above situation, a four-way parallel design of Starlink downlink signal frame synchronization lightweight circuit based on frequency offset correction assistance was proposed. Firstly, a lightweight coarse frame synchronization and frequency offset estimation structure based on delay correlation multiplexing was designed, so that the circuit would not need to calculate the delay correlation value again during frequency offset estimation. Then, a lightweight frequency offset correction module based on four-channel parallel DDS was proposed to avoid the resource consumption of buffering when a single-channel signal was converted under the multi-channel structure. Finally, the lightweight fine frame synchronization structure based on symbolic correlation and look-up table complex multiplier was designed to reduce the resource consumption of local sequence and signal, and the resources of LUT, LUTRAM, FF and BRAM were saved by 7%, 2%, 5% and 8% respectively under the premise of ensuring performance. Based on the Starlink open leading structure, a lightweight circuit design of Starlink downlink signal frame synchronization based on frequency offset correction assistance was proposed, and verilog code was written for FPGA implementation, and its resource occupation and performance were verified by the xczu47dr chip-on-board produced by Xilinx.关键词：Starlink;frame synchronization;lightweight55|66|0更新时间：2026-03-05
Research and Development
- 摘要：DeepSeek’s technological innovation has disrupted the high-investment model of traditional large models, triggering a transformation in the concept of computing power development and thus exerting an undeniable impact on the technical infrastructure architecture of data centers. From the perspective of demand traction, the impact of DeepSeek’s technological innovation on computing power demand was systematically analyzed. Based on this, the evolution trend of data center infrastructure was clarified, and a hierarchical solution of centralized nodes and distributed nodes as well as future development suggestions were proposed. The research conclusion provides theoretical references and practical guidance for the planning, construction and operation parties of data center infrastructure to better adapt to the development of artificial intelligence technology and build data center infrastructure in a forward-looking manner.关键词：intelligent computing reasoning;data center infrastructure;engineering productization;product engineering;intelligent operation and maintenance68|19|0更新时间：2026-03-05
- 摘要：How to ensure that network investment is directed to high return areas, how to achieve flexible deployment in phases, and how to quickly establish differentiated network advantages are key challenges faced by new entrants. To systematically address the issues, a wireless network planning method that integrated multi-source data and machine learning was proposed. Firstly, based on multi-source data sources, a multidimensional factor system was constructed, and the weights of each factor were optimized through machine learning model training to form a regional comprehensive scoring system. Then a multi-quadrant matrix strategy was used to adjust regional priorities and target population coverage rate, and a regional deployment priority list was generated. Finally, the method was verified and applied through actual case. It shows that this method is more ROI-oriented than traditional planning methods, which can effectively improve both network coverage rate and ROI. This method can effectively increase the commercial area coverage rate by 14%, 16% and 17% at three typical timeline in the 3rd, 5th and 10th years respectively, and increase the investment return rate by 21%, 19% and 16% respectively.关键词：network planning;multi-source data;greenfield deployment;machine learning;big data79|111|0更新时间：2026-03-05
- 摘要：In the 5G/6G era, owing to the widespread interconnection of high‑speed mobile communication networks and the Internet of things, network environments are rendered more open and complex, and the severity of diverse network intrusion threats is heightened. Traditional intrusion detection systems are predominantly based on centralized architectures, under which data from each domain are required to be aggregated at a central site for analysis; as a result, risks of data silos and privacy leakage are introduced, and adaptation to diverse threats across operator domains and networks is hindered. To address these issues, a federated learning based cross‑domain intrusion detection system model framework (Cross‑FL‑IDS) was proposed, in which intrusion detection models were trained locally within each network domain and model parameters were globally aggregated and updated, by which collaborative detection of emerging threats across domains was achieved. Under the premise that the privacy of each domain’s data was preserved, cross‑domain feature‑sharing and personalized fusion mechanisms were introduced in Cross‑FL‑IDS, through which the model’s generalization to heterogeneous traffic patterns was improved.关键词：federated learning;network security;intrusion detection;cross-domain collaboration61|66|0更新时间：2026-03-05
- 摘要：The establishment of a digital twin system for communication equipment rooms holds substantial significance in elevating the management level of communication networks and assets. Central to this endeavor is the low-cost yet high-quality three-dimensional modeling of the equipment and facilities within the rooms. For this reason, an intelligent 3D modeling framework for telecommunication equipment rooms was devised, whereby, leveraging multi-angle photographs captured by ordinary cameras and integrating advanced artificial intelligence (AI) technology, high-precision, semantically rich three-dimensional models of the equipment and facilities could be generated. This solution incorporated the structure from motion (SfM) and segmenting objects by locations (SOLO) algorithms, with the loss function of the SOLO algorithm being optimized. Experimental results demonstrate that this approach markedly enhances recognition accuracy, substantially boosts modeling efficiency, and effectively reduces the requisite quantity and precision threshold of photographs, thereby manifesting robust practicality.关键词：digital twin;3D modeling;SfM;SOLO;instance segmentation61|38|0更新时间：2026-03-05
Engineering and Application
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
- Total Visits：1035510 Visits Today：3559
0