Advanced Search
Articles in press have been peer-reviewed and accepted, which are not yet assigned to volumes /issues, but are citable by Digital Object Identifier (DOI).
Display Method:
Deepfake Video Detection on Social Networks Using Multi-domain Aware Driven by Common Mechanism Analysis Between Artifacts
WANG Yan, SUN Qindong, RONG Dongzhu, WANG Xiaoxiong
 doi: 10.11999/JEIT240025
[Abstract](102) [FullText HTML](30) [PDF 4707KB](18)
Abstract:
The misuse of deepfake technology on social networks has raised serious concerns about the authenticity and reliability of visual content. The degradation phenomenon of deepfake videos on social networks has not been adequately considered in existing detection algorithms, resulting in deepfake detection performance being limited by challenging issues such as compression artifacts interference and lack of context-related information. Compression encoding and up-sampling operations in deepfake generation algorithms can leave artifacts on videos, which can result in fine-grained differences between real videos and deepfake videos. The common mechanisms between compression artifacts and deepfake artifacts are analyzed to reveal the structural similarities between them, which provides a reliable theoretical basis for enhancing the robustness of deepfake detection models against compression. Firstly, to address the interference of compression noise on deepfake features, the frequency-domain adaptive notch filter is designed based on the structural similarity of compression artifacts and deepfake artifacts to eliminate the interference of compression artifacts on specific frequency bands. Secondly, the denoising branch based on residual learning is designed to reduce the sensitivity of the deepfake detection model to unknown noise. Additionally, the attention-based feature fusion method is adopted to enhance the discriminative features of deepfakes. Metric learning strategies are adopted to optimize network models, achieving deepfake detection with resistance to compression. Theoretical analysis and experimental results indicate that the detection performance of compressed deepfake videos is significantly enhanced by using the algorithm proposed in this paper. It can be used as a plug-and-play model combined with existing detection methods to enhance their robustness against compression.
Research on Constant False Alarm Rate Detection Technique for Ship in SAR Image
MENG Xiangwei
 doi: 10.11999/JEIT231436
[Abstract](133) [FullText HTML](158) [PDF 4143KB](93)
Abstract:
Among various methods to detect the ship targets in Synthetic Aperture Radar (SAR) image, the Constant False Alarm Rate (CFAR) detection algorithm with an adaptive detection threshold is the most important and extensively used one. In order to improve the detection performance for ships in SAR image, various statistical distributions are applied, with an attempt to accurately model the SAR clutter backgrounds, such as Gamma, K, log-normal, G0, the alpha-stable distribution, etc. In modern radar systems, the use of the CFAR technique is necessary to keep the false alarms at a suitably low rate in an a priori unknown time-varying and spatially nonhomogeneous backgrounds, and to improve the detection probability as much as possible. The clutter background in SAR images is complicated and variable, when the actual clutter background deviates from the assumed statistical distribution, the performance of the parametric CFAR detectors deteriorates, whereas the nonparametric CFAR method exhibits its advantage. In this paper, the Wilcoxon nonparametric CFAR scheme for ship detection in SAR image is proposed and analyzed. By comparison with several typical parametric CFAR schemes on 3 real SAR images of Radarsat-2, ICEYE-X6 and Gaofen-3, the robustness of the Wilcoxon nonparametric detector to maintain a good false alarm performance in these different detection backgrounds is revealed, and its detection performance for the weak ship is improved evidently. Moreover, the detection speed of the Wilcoxon nonparametric detector is fast, and it has the simplicity of hardware implementation.
Method of Maximizing Sum Rate for Dual STAR-RIS Assisted Downlink NOMA Systems
TIAN Xinji, MENG Haoran, LI Xingwang, ZHANG Hui
 doi: 10.11999/JEIT240007
[Abstract](61) [FullText HTML](20) [PDF 1619KB](18)
Abstract:
A sum-rate maximization method is proposed for two Simultaneously Transmitting And Reflecting Reconfigurable Intelligent Surface (STAR-RIS) assisted downlink Non-Orthogonal Multiple Access (NOMA) systems. Firstly, the optimization problem for maximizing sum rate is constructed, with STAR-RIS phase shifts, power allocation and time allocation as optimization parameters. Then the Semi-Definite Programming method (SDP) is used to optimize the phase shifts of these two STAR-RISs. Finally, the power allocation and time allocation are optimized alternately by iterative method. In each iteration, Lagrange dual decomposition method is used to optimize power allocation and function extremum method is used to optimize time allocation. Simulation results show that the sum rate of the dual STAR-RIS-assisted NOMA system is higher than that of the single STAR-RIS-assisted NOMA system.
Secret Sharing: Design of Higher-Order Masking S-box and Secure Multiplication in Galois Field
TANG Xiaolin, FENG Yan, LI Mingda, LI Zhiqiang
 doi: 10.11999/JEIT231272
[Abstract](37) [FullText HTML](15) [PDF 4473KB](4)
Abstract:
In the information era, information security is the priority that cannot be ignored. Attacks and protection against password devices are research hotspots in this field. In recent years, various attacks on cryptographic devices have become well-known, all aimed at obtaining keys from the device. Among these attacks, power side channel attack is one of the most concerned attack techniques. Mask technology is an effective method to combat power side channel attacks, however, with the continuous progress of attack methods, the protection of first-order mask is no longer sufficient to cope with second-order and higher order power analysis attack, so the research on higher-order mask has considerable significance. To enhance the encryption circuit’s capability of anti-attack, high-order masking schemes: N-share masking is implemented on S-box in this paper, and a universal design method for galois field secure multiplication is proposed, which is based on the secure scheme published by Ishai et al. at Crypto 2003 (ISW framework). Through experiments, it has been shown that the encryption scheme adopted in this paper does not affect the functionality of the encryption algorithm, and can resist first-order and second-order correlation power analysis attack.
Elevation Error Prediction Dataset Using Global Open-source Digital Elevation Model
YU Cuilin, WANG Qingsong, ZHONG Zixuan, ZHANG Junhao, LAI Tao, HUANG Haifeng
 doi: 10.11999/JEIT240062
[Abstract](72) [FullText HTML](20) [PDF 7142KB](15)
Abstract:
The correction in Digital Elevation Models (DEMs) has always been a crucial aspect of remote sensing geoscience research. The burgeoning development of new machine learning methods in recent years has provided novel solutions for the correction of DEM elevation errors. Given the reliance of machine learning and other artificial intelligence methods on extensive training data, and considering the current lack of publicly available, unified, large-scale, and standardized multisource DEM elevation error prediction datasets for large areas, the multi-source DEM Elevation Error Prediction Dataset (DEEP-Dataset) is introduced in this paper. This dataset comprises four sub-datasets, based on the TerraSAR-X add-on for Digital Elevation Measurements (TanDEM-X) DEM and Advanced land observing satellite World 3D-30 m (AW3D30) DEM in the Guangdong Province study area of China, and the Shuttle Radar Topography Mission (SRTM) DEM and Advanced Spaceborne Thermal Emission and reflection Radiometer (ASTER) DEM in the Northern Territory study area of Australia. The Guangdong Province sample comprises approximately 40 000 instances, while the Northern Territory sample includes about 1 600 000 instances. Each sample in the dataset consists of ten features, encompassing geographic spatial information, land cover types, and topographic attributes. The effectiveness of the DEEP-Dataset in actual model training and DEM correction has been validated through a series of comparative experiments, including machine learning model testing, DEM correction, and feature importance assessment. These experiments demonstrate the dataset’s rationality, effectiveness, and comprehensiveness.
Scale Adaptive Fusion Network for Multimodal Remote Sensing Data Classification
LIU Xiaomin, YU Mengjun, QIAO Zhenzhuang, WANG Haoyu, XING Changda
 doi: 10.11999/JEIT240178
[Abstract](32) [FullText HTML](23) [PDF 3321KB](9)
Abstract:
The multimodal fusion method can effectively improve the ground object classification accuracy by using the complementary characteristics of different modalities, which has become a research hotspot in the various fields in recent years. The existing multimodal fusion methods have been successfully applied to multi-source remote sensing classification tasks oriented to HyperSpectral Image (HSI) and Light Detection And Ranging (LiDAR). However, existing research still faces many challenges, including difficulty in capturing spatial dependencies among irregular ground objects and obtaining discriminative information in multimodal data. To address the above challenges, a Scale Adaptive Fusion Network (SAFN)is proposed in this paper, by integrating the fusion of multimodal, multiscale, and multiview features into a unified framework. First, a dynamic multiscale graph module is proposed to capture the complex spatial dependencies of ground object, enhancing the model’s adaptability to irregular and scale-dissimilar ground object. Second, the complementary properties of LiDAR and HSI are utilized to constrain ground object within the same spatial neighborhood to have similar feature representations, thereby acquiring discriminative remote sensing features. Then, a multimodal spatial-spectral graph fusion module is proposed to establish feature interactions among multimodal, multiscale, and multiview features, providing discriminative fusion features for classification tasks by capturing class-recognition information that can be shared among features. Finally, the fusion features are fed into a classifier to obtain class probability scores for predicting the ground object class. To verify the effectiveness of SAFN, experiments are conducted on three datasets (i.e., Houston, Trento, and MUUFL). The experimental results show that, SAFN achieved state-of-the-art performance in multi-source remote sensing data classification tasks when compared with existing mainstream methods.
Lightweight Self-supervised Monocular Depth Estimation Method with Enhanced Direction-aware
CHENG Deqiang, XU Shuai, LÜ Chen, HAN Chengong, JIANG He, KOU Qiqi
 doi: 10.11999/JEIT240189
[Abstract](34) [FullText HTML](16) [PDF 7196KB](8)
Abstract:
To address challenges such as high complexity in monocular depth estimation networks and low accuracy in regions with weak textures, a Direction-Aware Enhancement-based lightweight self-supervised monocular depth estimation Network (DAEN) is proposed in this paper. Firstly, the Iterative Dilated Convolution module (IDC) is introduced as the core of the encoder to extract correlations among distant pixels. Secondly, the Directional Awareness Enhancement module (DAE) is designed to enhance feature extraction in the vertical direction, providing the depth estimation model with additional depth cues. Furthermore, the problem of detail loss during the decoder upsampling process is addressed through the aggregation of disparity map features. Lastly, the Feature Attention Module (FAM) is employed to connect the encoder and decoder, effectively leveraging global contextual information to resolve adaptability issues in regions with weak textures. Experimental results on the KITTI dataset demonstrate that the proposed method has a model parameter count of only 2.9M, achieving an advanced performance with \begin{document}$ \delta $\end{document} metric of 89.2%. The generalization of DAEN is validated on the Make3D datasets, with results indicating that the proposed method outperforms current state-of-the-art methods across various metrics, particularly exhibiting superior depth prediction performance in regions with weak textures.
Airborne Target Tracking Algorithm Using Multi-Platform Heterogeneous Information Fusion
PENG Ruihui, GUO Wei, SUN Dianxing, TAN Shuo, DOU Yuecong
 doi: 10.11999/JEIT240130
[Abstract](54) [FullText HTML](15) [PDF 8098KB](14)
Abstract:
An innovative aviation target tracking algorithm is presented in this paper, utilizing high-altitude unmanned airship dual photoelectric sensors in conjunction with Unmanned Aerial Vehicle (UAV)-borne two-coordinate radar. The algorithm addresses the challenge of integrating sensor data to accurately track targets when individual sensors lack complete target position information, thus overcoming limitations of traditional point-trace association methods. Initially, a two-level point-trace correlation algorithm based on angle and distance is introduced for multi-sensor measurement association following coordinate system transformation. Subsequently, a line-plane intersection fusion localization algorithm is proposed to determine the initial target track position through techniques such as least squares method, intersection projection, distance nearest point solution, and homologous data compression. Leveraging heterogeneous information from space-based multi-platform reconnaissance, an extended Unscented Kalman Filter (UKF) is designed to track aviation targets by enhancing the traditional UKF. Simulation results demonstrate that this algorithm achieves superior precision in tracking high-speed aerial targets.
Joint Internal and External Parameters Calibration of Optical Compound Eye Based on Random Noise Calibration Pattern
LI Dongsheng, WANG Guoyan, LIU Jinxin, FAN Hongqi, LI Biao
 doi: 10.11999/JEIT230652
[Abstract](104) [FullText HTML](30) [PDF 9742KB](12)
Abstract:
In tasks such as precise guidance and obstacle avoidance navigation based on optical compound eyes, the calibration of optical compound eyes plays a crucial role in achieving high accuracy. The classical Zhang's calibration method requires each ommatidium of the optical compound eyes to observe a complete chessboard pattern. However, the complexity of the optical compound eye structure makes it difficult to satisfy this requirement in practical applications. In this paper, a joint internal and external parameters calibration algorithm of optical compound eyes based on a random noise calibration pattern is proposed. This algorithm utilizes the local information captured by the ommatidia when photographing the random noise calibration pattern, enabling simple and fast calibration for optical compound eyes with arbitrary configurations and numbers of ommatidia. To improve the robustness of the calibration, a multi-threshold matching mechanism is introduced to address the issue of sparse feature point quantity in ommatidial visual fields leading to matching failures. Moreover, an error model for the joint internal and external parameters calibration of optical compound eyes is presented to evaluate the accuracy of the proposed algorithm. Experimental comparisons with Zhang’s calibration method demonstrate the robustness of the proposed algorithm. Furthermore, the high accuracy of the proposed joint calibration algorithm is validated in a physical system of optical compound eyes.
Research on Intelligent Spectrum Allocation Techniques for Incomplete Electromagnetic Information
ZHAO Haoqin, DUAN Guodong, SI Jiangbo, HUANG Rui, SHI Jia
 doi: 10.11999/JEIT231005
[Abstract](115) [FullText HTML](40) [PDF 5190KB](24)
Abstract:
To solve the problem of low spectrum utilization of multi-node autonomous frequency decision-making in the dynamic electromagnetic countermeasure environment, the research on intelligent cooperative spectrum allocation technology for non-complete electromagnetic information is carried out, which improves spectrum utilization through multi-node intelligent collaboration. Firstly, the spectrum allocation problem is modelled as an optimization problem to maximize the frequency-using equipment, and secondly, a resource decision-making algorithm based on the multi-node cooperative diversion experience repetition mechanism (Cooperation- Deep double Q-network, Co-DDQN) is proposed. This algorithm evaluates the historical experience data based on the cooperative diversion function and is trained by a hierarchical experience pool, so that each agent can form a lightweight cooperative decision-making ability under self-observation, and solve the problem of inconsistency between the optimization direction of multi-node decision-making and the overall optimization goal under low-visibility conditions. Besides, a hybrid reward function based on confidence allocation is designed, and each node considers itself when the decision is made, which can reduce the emergence of lazy nodes, explore a better overall action strategy, and further improve the system efficiency. Simulation results show that when the number of nodes is 20, the number of accessible devices of the proposed algorithm outperforms the global greedy algorithm and the genetic algorithm, and the difference with the centralized spectrum allocation algorithm with complete information sharing is within 5%, which is more suitable for cooperative spectrum allocation of low-visibility nodes.
Research on Energy Consumption Minimization for a Data Compression Based NOMA-MEC System
SHI Liqin, LIU Xuan, LU Guangyue
 doi: 10.11999/JEIT231033
[Abstract](125) [FullText HTML](51) [PDF 1861KB](18)
Abstract:
The system energy consumption minimization problem is studied for a data compression based Non-Orthogonal Multiple Access-Mobile Edge Computing (NOMA-MEC) system. Considering the partial compression and offloading schemes and the limited computation capacity at the base station, a system energy consumption minimization optimization problem is formulated by jointly optimizing the users’ data compression and offloading ratios, transmit power, data compression time, etc. In order to solve this problem, closed-form expression of each user’s optimal transmit power is firstly derived. Then the Successive Convex Approximation (SCA) method is used to approximate the non-convex constraints of the formulated problem, and An SCA based efficient iterative algorithm is proposed to solve the formulated problem, obtaining the optimal resource allocation scheme of the system. Finally, the simulation results verify the advantages of the proposed scheme via computer simulations and show that compared with other benchmark schemes, the proposed scheme can effectively reduce the system energy consumption.
Screen-Shooting Resilient Watermarking Scheme Combining Invertible Neural Network and Inverse Gradient Attention
LI Xiehua, LOU Qin, YANG Junxue, LIAO Xin
 doi: 10.11999/JEIT230953
[Abstract](227) [FullText HTML](90) [PDF 5415KB](38)
Abstract:
With the growing use of smart devices, the ease of sharing digital media content has been enhanced. Concerns have been raised about unauthorized access, particularly via screen shooting. In this paper, a novel end-to-end watermarking framework is proposed, employing invertible neural networks and inverse gradient attention, to tackle the copyright infringement challenges related to screen content leakage. A single invertible neural network is employed by the proposed method for watermark embedding and extraction, ensuring information integrity during network propagation. Additionally, robustness and visual quality are enhanced by an inverse gradient attention module, which emphasizes pixel values and embeds the watermark in imperceptible areas for better invisibility and model resilience. Model parameters are optimized using the Learnable Perceptual Image Patch Similarity (LPIPS) loss function, minimizing perception differences in watermarked images. The superiority of this approach over existing learning-based screen-shooting resilient watermarking methods in terms of robustness and visual quality is demonstrated by experimental results.
Self-tuning Multivariate Variational Mode Decomposition
LANG Xun, WANG Jiayi, CHEN Qiming, HE Bingbing, MAO Rukai, XIE Lei
 doi: 10.11999/JEIT230763
[Abstract](145) [FullText HTML](45) [PDF 5682KB](16)
Abstract:
The Multivariate Variational Mode Decomposition (MVMD), being an extension of the Variational Mode Decomposition (VMD), inherits the merits of VMD. However, it encounters an issue wherein its decomposition performance relies heavily on two predefined parameters, the number of modes (K) and the penalty factor (\begin{document}$ \alpha $\end{document}). To address this issue, a Self-tuning MVMD (SMVMD) algorithm is proposed. SMVMD employs the notion of matching pursuit to adaptively update K and \begin{document}$ \alpha $\end{document} based on energy occupation and mode orthogonality in the frequency domain, respectively. The experimented results of both simulated signals and real cases demonstrate that the proposed SMVMD not only effectively addresses the parameter rectification problem of the original MVMD, but also exhibits the following advantages: (i) SMVMD displays superior resilience to mode-mixing compared to MVMD, along with enhanced robustness to both noise and variations in \begin{document}$ \alpha $\end{document}-value. (ii) In comparison to the classical algorithms of multivariate empirical mode decomposition, fast multivariate empirical mode decomposition, and multivariate variational mode decomposition, SMVMD showcases the lowest decomposition error and the best decomposition effect.
Integrating Multiple Context and Hybrid Interaction for Salient Object Detection
XIA Chenxing, CHEN Xinyu, SUN Yanguang, GE Bin, FANG Xianjin, GAO Xiuju, ZHANG Yan
 doi: 10.11999/JEIT230719
[Abstract](253) [FullText HTML](88) [PDF 5953KB](53)
Abstract:
Salient Object Detection (SOD) aims to recognize and segment visual salient objects in images, which is one of the important research contents in computer vision tasks and related fields. Existing Fully Convolutional Networks (FCNs)-based SOD methods have achieved good performance. However, the types and sizes of salient objects are variable and unfixed in real-world scenes, which makes it still a huge challenge to detect and segment salient objects accurately and completely. For that, in this paper, a novel integrating multiple context and hybrid interaction for SOD task is proposed to efficiently predict salient objects by collaborating Dense Context Information Exploration (DCIE) module and Multi-source Feature Hybrid Interaction (MFHI) module. The DCIE module uses dilated convolution, asymmetric convolution and dense guided connection to progressively capture the strongly correlated multi-scale and multi-receptive field context information, and enhances the expression ability of each initial input feature by aggregating context information. The MFHI module contains diverse feature aggregation operations, which can adaptively interact with complementary information from multi-level features to generate high-quality feature representations for accurately predicting saliency maps. The performance of the proposed method is tested on five public datasets. The performance of the proposed method is tested on five public datasets. Experimental results demonstrate that our method achieves superior prediction performance compared with 19 state-of-the-art SOD methods under different evaluation metrics.
Intelligent Weighted Energy Consumption and Delay Optimization for UAV-Assisted MEC Under Malicious Jamming
YANG Helin, ZHENG Mengting, LIU Shuai, XIAO Liang, XIE Xianzhong, XIONG Zehui
 doi: 10.11999/JEIT230986
[Abstract](160) [FullText HTML](54) [PDF 3867KB](27)
Abstract:
In recent years, mounting Mobile Edge Computing (MEC) servers on Unmanned Aerial Vehicle (UAV) to provide services for mobile ground users has been widely researched in academia and industry. However, in malicious jamming environments, how to effectively schedule resources to reduce system delay and energy consumption becomes a key challenge. Therefore, this paper considers a UAV-assisted MEC system under a malicious jammer, where an optimization model is established to minimize the weighted energy consumption and delay by jointly optimizing UAV flight trajectories, resource scheduling, and task allocation. As the optimization problem is difficult to be solved and the malicious jamming behavior is dynamic, a Twin Delayed Deep Deterministic (TD3) policy gradient algorithm is proposed to search for the optimal policy. At the same time, the Prioritized Experience Replay (PER) technique is added to improve the convergence speed and stability of the algorithm, which is highly effective against malicious interference attacks. The simulation results show that the proposed algorithm can effectively reduce the delay and energy consumption, and achieve good convergence and stability compared with other algorithms.
2024, 46(6): 1-1.  
[Abstract](22) [FullText HTML](9) [PDF 19021KB](6)
Abstract:
2024, 46(6): 1-4.  
[Abstract](24) [FullText HTML](12) [PDF 279KB](6)
Abstract:
Dataset
ShuangQing-1 (Luojia3-01) Multimode Imaging Sample Dataset
WANG Mi, YANG Fang, LI Deren, PAN Jun, DAI Rongfan
2024, 46(6): 2299-2310.   doi: 10.11999/JEIT230921
[Abstract](291) [FullText HTML](309) [PDF 14483KB](61)
Abstract:
Herein, the Shuangqing-1(Luojia3-01) multimode imaging sample dataset is presented to address the problem of limited data types provided for user services by remote sensing satellites with the highest resolution. This dataset includes various imaging modes, such as push-scan, array push-frame, and video staring; hence, it covers typical data samples from different target areas ,such as urban regions, water bodies, mountainous regions, and airports. The construction of this dataset involves signal data decoding, Bayer interpolation, relative radiometric correction, geometric positioning, video stabilization, and three-dimensional reconstruction. Additionally, in-depth discussions and investigations are conducted on key algorithms, such as on-orbit calibration, rapid production of area of interest products, high-definition video geometric stabilization, and multi-angle three-dimensional reconstruction. Finally, the sample dataset is visually displayed and quantitatively evaluated from three aspects: image standard, video staring, and real-world three-dimensional products.
Overviews
Key Technologies and Development Trends of Free Space Optical UAV Communication Network
FENG Simeng, ZHAO Yidi, DONG Chao, WU Qihui
2024, 46(6): 2311-2322.   doi: 10.11999/JEIT230644
[Abstract](478) [FullText HTML](298) [PDF 1567KB](87)
Abstract:
Considering the electromagnetic spectrum congestion and serious interference, the Free Space Optical (FSO)-based Unmanned Aerial Vehicle (UAV) communication network constitutes an important part for the space-air-ground integration, attracting substantial attention from both academia and industry. Compared to radio frequency communication, FSO communication is benefited from high data rate, low latency and high security. However, the FSO link is susceptible to atmospheric environment, while the mobile UAV dynamics topology and limited resources bring further challenges. Therefore, this paper first introduces the FSO transmission characteristics and then focuses on the key technologies to enhance stability and quality of FSO-based UAV networks. Furthermore, the development trend of FSO-based UAV network, in terms of high reliability, strong intelligence and long endurance is analyzed.
An Overview of Novel Multi-access Techniques for Multi-dimensional Expanded 6G
PANG Xiaowei, JIANG Xu, LU Huabing, ZHAO Nan
2024, 46(6): 2323-2334.   doi: 10.11999/JEIT231265
[Abstract](569) [FullText HTML](161) [PDF 2502KB](108)
Abstract:
With the evolution of mobile communication technology, the Sixth-Generation (6G) wireless networks will achieve a leap from the internet of things to the internet of intelligent things, meeting higher data demands and broader application scenarios. Novel multiple access technologies and multidimensional expansion techniques will jointly play a role in 6G, providing crucial support for building an efficient, intelligent, and reliable communication network to meet the diverse demands of future communications. Therefore, this review paper aims to explore the application potentials of novel multiple access technologies in multidimensional expansion 6G communication networks. Firstly, it compares traditional multiple access technologies with potential novel multiple access technologies in 6G, with a focus on the advantages of non-orthogonal multiple access technology in improving spectral efficiency and system capacity. Then, it provides a detailed introduction to the advantages and functions of multidimensional expansion technologies such as satellite communication, Unmanned Aerial Vehicle (UAV), and Intelligent Reflecting Surface (IRS) in 6G scenarios. Furthermore, the advantages and collaborative applications of novel multiple access technologies in conjunction with satellite communication, UAV, and IRS are discussed. Finally, the paper discusses key technological challenges in a novel multi-dimensional extension network based on new multiple access technologies, including large-scale multiple-input-multiple-output, terahertz technology, integrated sensing, communication, and computing, user information security, and imperfect Channel State Information (CSI) estimation, while also providing prospects for new coding technologies, artificial intelligence and machine learning.
Review on Olfactory and Visual Neural Pathways in Drosophila
ZHANG Sheng, ZHENG ShengNan, SHEN Jie, YIN Xinghui, XU Lizhong
2024, 46(6): 2335-2351.   doi: 10.11999/JEIT230508
[Abstract](131) [FullText HTML](89) [PDF 16612KB](32)
Abstract:
The olfactory and visual neural systems in Drosophila are highly sensitive to the olfactory and visual stimuli in the natural environment. The highly sensitive single-modal perception and cross-modal collaboration decision-making mechanisms of the olfactory and visual neural systems provide certain inspiration for bionic applications. Firstly, based on the olfactory and visual neural systems in Drosophila, the current research status of the physiological mechanisms and computational models of single-modal perception decision-making of the olfactory and visual neurons is summarized. The summary is divided into three parts: capturing, processing, and decision-making of the olfactory and visual signals. Meanwhile, the physiological mechanisms and computational models of cross-modal collaboration decision-making of the olfactory and visual neurons in Drosophila are further expounded. Then, the typical bionic applications of single-modal perception and cross-modal collaboration in Drosophila are summarized. Finally, the current challenges of the physiological mechanisms and computational models of the olfactory and visual neural pathways in Drosophila are summarized and the future development trends are outlooked for, which lays a foundation for future research work.
Wireless Communication,Internet of Things and Digital Signal Processing
Consistent-coverage Oriented AP Deployment Optimization in Cell Free and Legacy Coexistence Network
JIANG Jing, TAO Sha, WANG Wei, CHU Hongyun, Worakrin Sutthiphan, LI Chunguo
2024, 46(6): 2352-2360.   doi: 10.11999/JEIT230627
[Abstract](170) [FullText HTML](104) [PDF 2688KB](32)
Abstract:
To address the issue of dramatic fluctuations in user experience in legacy cellular networks, cell-free and legacy coexistence networks deploy Access Points (APs) into cellular networks, which can significantly improve the coverage signal quality of edge users and blind areas. Therefore, a good and consistent user experience at any location in the coverage area, i.e. consistent-coverage is the primary goal to improve the performance of coexistence networks. As the AP deployment is the determinant of user transmission rate and coverage in coexistence networks, a consistent-coverage oriented AP deployment optimization problem is designed. Firstly, the expression of the downlink achievable rate of each user is derived based on the joint transmission model of coexistence network. Secondly, a ratio sum optimization problem is proposed to maximize the average throughput. Finally, the non-convex problem is transformed into a convex optimization problem by using the fractional programming and the introduction of auxiliary variables, where the AP deployment scheme is obtained by the iterative solution. Compared with the legacy cellular networks, the simulation results demonstrate that the proposed scheme can significantly increase average throughput of the edge and blind areas.
Proximal Policy Optimization Algorithm for UAV-assisted MEC Vehicle Task Offloading and Power Control
TAN Guoping, Yi Wenxiong, ZHOU Siyuan, HU Hexuan
2024, 46(6): 2361-2371.   doi: 10.11999/JEIT230770
[Abstract](186) [FullText HTML](113) [PDF 2443KB](41)
Abstract:
The architecture of Mobile Edge Computing (MEC), assisted by Unmanned Aerial Vehicles (UAVs), is an efficient model for flexible management of mobile computing-intensive and delay-sensitive tasks. Nevertheless, achieving an optimal balance between task latency and energy consumption during task processing has been a challenging issue in vehicular communication applications. To tackle this problem, this paper introduces a model for optimizing task offloading and power control in vehicle networks based on UAV-assisted mobile edge computing architecture, using a Non-Orthogonal Multiple Access (NOMA) approach. The proposed model takes into account dynamic factors like vehicle high mobility and wireless channel time-variations. The problem is modeled as a Markov decision process. A distributed deep reinforcement learning algorithm based on Proximal Policy Optimization (PPO) is proposed, enabling each vehicle to make autonomous decisions on task offloading and related transmission power based on its own perceptual local information. This achieves the optimal balance between task latency and energy consumption. Simulation results reveal that the proposed proximal policy optimization algorithm for task offloading and power control scheme not only improves the performance of task latency and energy consumption compared to existing methods, The average system cost performance improvement is at least 13% or more. but also offers a performance-balanced optimization method. This method achieves optimal balance between the system task latency and energy consumption level by adjusting user preference weight factors.
Backscatter-NOMA Enabled Hybrid Multicast-Unicast Cooperative Transmission Scheme
KUO Yonghong, XUE Yanwen, LÜ Lu, HE Bingtao, CHEN Jian
2024, 46(6): 2372-2381.   doi: 10.11999/JEIT230672
[Abstract](81) [FullText HTML](39) [PDF 2544KB](11)
Abstract:
In order to address the low spectral efficiency and inefficient link utilization problem in cooperative relay communication system, a Backscatter-NOMA enabled hybrid multicast-unicast cooperative transmission scheme is proposed for the scenario of coexistence of multicast and unicast services. A multicast user is opportunistically selected as a cooperative node, which used a part of the power of the received signal for its own decoding, and backscatter the residual power to enhance the reception quality of other users. To improve system performance, the minimum achievable rate for unicast users is maximized by jointly optimizing the base station power allocation coefficients, cooperative user backscatter coefficients and cooperative node selection variables, while guaranteeing the quality of service for multicast. To solve the above highly non-convex joint optimization problem, a cooperative user selection criterion was designed and an iterative algorithm was proposed to obtain the optimal solution to the original problem. The simulation results verify the fast convergence of the proposed iterative algorithm, which can improve the minimum achievable rate of unicast users by 11.5% compared to the non-cooperative transmission scheme, and effectively ensure the quality of multi-service.
Joint Trajectory and Resource Allocation Optimization for Air-ground Collaborative Integrated Sensing and Communication Systems
ZHANG Guangchi, GU Zelin, CUI Miao
2024, 46(6): 2382-2390.   doi: 10.11999/JEIT230716
[Abstract](223) [FullText HTML](117) [PDF 3032KB](60)
Abstract:
An air-ground collaborative integrated sensing and communication system is studied, where the air-ground collaborative network is composed of an Unmanned Ground Vehicle (UGV) base station and Unmanned Aerial Vehicle (UAV) relays. The network provides communication service for ground users while detecting and sensing target areas. The air-ground channels are modeled as the accurate Rician fading channel model. On the constraints of the sensing frequency and the effective sensing power threshold of the target areas, the minimum average communication rate of all users is maximized by jointly optimizing the communication and sensing association of the system, the transmit power and flight trajectory of the UAV relays, as well as the transmit power and trajectory of the UGV base station. To solve the resultant non-convex integer optimization problem with highly coupled variables, the block coordinate descent method is applied to decompose the original optimization problem into four sub-problems, where relaxation variables are introduced, and the integer constraints are converted into penalty terms. Then, it is proved that the effective sensing power is a composition function of the trajectory variables and the relaxation variables and is a jointly convex function of them, so that the non-convex terms are tackled by using the successive convex optimization method. Lastly, a two-layer iterative algorithm is proposed to obtain the suboptimal solution efficiently. It is shown by simulation results that compared to some benchmark algorithms, the proposed algorithm significantly increases the minimum average communication rate of all users while achieving the same sensing performance and achieves a better performance trade-off between communication and sensing with good convergence performance.
Channel State Information Restoration and Positioning of massive Multiple Input Multiple Output Integrated Visible Light Communication and Sensing System
LIU Xiaodong, NING Yiting, DONG Fan, TANG Liwei, WANG Yuhao, WANG Jinyuan
2024, 46(6): 2391-2400.   doi: 10.11999/JEIT231389
[Abstract](142) [FullText HTML](97) [PDF 3055KB](30)
Abstract:
Benefiting from rich spectrum and lamps, Integrated Visible Light Communication and Positioning (IVLCP) systems provide powerful technological solution to meet the high performance communication and positioning in indoor wireless networks. Meanwhile, the massive Multiple Input Multiple Output (m-MIMO) effectively enhance both service coverage and quality of IVLCP systems. However, the channel environment is more complex and the priori information rapidly changed in the m-MIMO-enabled IVLCP systems, making traditional methods challenging for fast and accurate channel estimation and positioning. In order to tackle this challenge, a Channel State Information Restoration and Positioning (CSIRP) network is proposed in this paper. The network not only effectively captures complex distribution feature of channel but also addresses the temporal variations in location, thereby enhancing the robustness and dynamic adaptability of channel and location estimation. Specifically, the CSIRP network employs a conditional generative adversarial process to adaptively train the generator and discriminatorr and thus achieves the channel estimation from the received signals. Then, the Long Short-Term Memory(LSTM) is introduced to estimate the location of the receiver from the estimated channel. Simulation results demonstrate that the accuracy of both channel and location estimation achieved by the proposed CSIRP network outperforms existing deep learning benchmark schemes. This provides m-MIMO-enabled IVLCP systems with more reliable and accurate channel state information and positioning.
Outage Performance of Tag Selection Scheme for Backscatter Communication Systems
LIU Yingting, ZHOU Zhiyang, GENG Mengdan, LI Xingwang
2024, 46(6): 2401-2408.   doi: 10.11999/JEIT231001
[Abstract](158) [FullText HTML](53) [PDF 2093KB](18)
Abstract:
The considered Backscatter Communication (BackCom) system consists of one dedicated radio frequency source node, some tags and one destination node. In consideration of the Channel Estimation Error (CEE), the tag selection scheme in which the tag selection scheme that can maximize the received Signal-to-Noise Ratio (SNR) at the destination is proposed over the Nakagami-m channels, and the corresponding analytical results of the outage probability and diversity gain are derived. In this paper, the consumed power by tags is considered. The numerical results verify the obtained analytical results and investigate the key parameters on the system performance. Both the analytical and numerical results show that the existence of the CEE make the corresponding diversity gain zero.
Research on Multi-User Detection Algorithm for Non-Orthogonal Multiple Access Short Message Based on Low Complexity Adder Network
WANG Ji, LI Zilong, XIAO Jian, LI Huanzhe, XIE Wenwu, YU Chao
2024, 46(6): 2409-2417.   doi: 10.11999/JEIT231186
[Abstract](78) [FullText HTML](29) [PDF 3277KB](13)
Abstract:
A joint constellation trace diagram and deep learning-based blind modulation detection scheme is proposed for Non-Orthogonal Multiple Access (NOMA) systems, which can avoid the required expensive signaling overhead in successive interference cancellation algorithms, especially for NOMA-based short packet transmission. Considering the high computational complexity and energy consumption for communication equipment in the deployment of neural network, the original convolutional network is replaced by the adder network. The modulation detection accuracy, computing delay and energy consumption are fully compared for two kinds of network architectures. Meanwhile, time-domain oversampling technology is used to improve the recognition rate under low signal-to-noise ratio. Finally, the influence of power allocation and data packet length on detection performance is analyzed and verified.
Low Complexity Receiver Design for Orthogonal Time Frequency Space Systems
LIAO Yong, LI Xue
2024, 46(6): 2418-2424.   doi: 10.11999/JEIT230625
[Abstract](301) [FullText HTML](97) [PDF 2496KB](27)
Abstract:
Orthogonal Time Frequency Space (OTFS) can convert the doubly-selective channels into non-selective channels in the Delay-Doppler (DD) domain, which provides a solution for establishing reliable wireless communication in high-mobility scenarios. However, serious Inter-Doppler Interference (IDI) exists in complex multi-scattering scenarios such as internet of vehicles, which brings great challenges to the accurate demodulation of OTFS receiver signals. To solve these problems, a kind of joint Sparse Bayesian Learning (SBL) and damped Least Square Minimum Residual (d-LSMR) OTFS receiver is proposed. Firstly, based on the relationship between OTFS time domain and DD domain, the channel estimation problem is transformed into a Basis Expansion Model (BEM) to accurately estimate DD domain channels including Doppler sampling points. Then, an efficient conversion algorithm is proposed to convert the basis coefficients into channel equivalent matrix. Additionally, the noise estimated in channel estimation is used in d-LSMR equalizer, and the sparse channel matrix in DD domain is adopted to achieve fast convergence. System simulation results show that compared with the current representative OTFS receiver, the proposed scheme achieves better bit error rate performance and reduces the computational complexity.
Diffuse Scattering Propagation and Depolarization Modeling for B5G Millimeter-wave Communications at 40~50 GHz
LIAO Xi, CHEN Xinrui, WANG Yang, REN Minghao, CHEN Qianbin
2024, 46(6): 2425-2433.   doi: 10.11999/JEIT230706
[Abstract](96) [FullText HTML](29) [PDF 7835KB](11)
Abstract:
To achieve an accurate characterization and thorough understanding of the millimeter-Wave (mmWave) communication channel propagation mechanism, precise characterization of diffuse scattering propagation and polarization is crucial. These aspects are also indispensable for the development of a high-precision mmWave channel model. In response to the insufficient characterization of diffuse scattering propagation and polarization caused by building materials in the mmWave frequency bands, a novel diffuse scattering depolarization modeling method based on effective roughness theory is presented in this paper. Initially, the electric field of the diffuse scattering radiation produced by the rough surface of building materials is decomposed along the polarization dimension of the electromagnetic wave. Subsequently, the depolarization coefficient is introduced to establish the propagation model. The paper investigates the diffuse scattering propagation and polarization characteristics, encompassing the power angle spectrum, depolarization coefficient, and cross-polarization discrimination ratio, by utilizing measured data of typical materials in the 40~50 GHz frequency range. The obtained results validate that the proposed model effectively captures the polarization characteristics of building materials with both rough and smooth surfaces. The achieved depolarization conversion rates are 39% and 4%, respectively.
Communication and Sensing Performance Analysis of RIS-assisted SWIPT-NOMA System under Non-ideal Conditions
LI Xingwang, WANG Xinying, TIAN Xinji, WANG Xinshui, QIN Panke, CHEN Hui
2024, 46(6): 2434-2442.   doi: 10.11999/JEIT231395
[Abstract](176) [FullText HTML](44) [PDF 3554KB](27)
Abstract:
To meet the escalating demands for efficient communication and reliable sensing, a Reconfigurable Intelligent Surface(RIS)-assisted Simultaneous Wireless Information and Power Transfer(SWIPT)-Non-Orthogonal Multiple Access(NOMA) system is proposed in this paper. This system is designed to concurrently achieve target sensing and information transmission. Considering the imperfections of Successive Interference Cancellation (SIC) and Channel Estimation Error (CEE), a comprehensive analysis of the system’s reliability, effectiveness, and radar sensing performance is conducted. Analytical expressions for the Outage Probability (OP), Ergodic Rate (ER), Probability of Detection (PoD), and Radar Estimation Information Rate (REIR) of the system are derived to provide insights into its performance. The analysis results reveal the following findings: the presence of imperfect SIC and CEE adversely impacts the system’s performance; the OP diminishes as the transmitted power of base station increases, eventually converging to a constant value in the high Signal-to-Noise Ratio (SNR) region; both the ER and the REIR increase with the base station’s transmitted power and eventually stabilize at an upper limit value in the high SNR region; the PoD increases with the base station’s transmit power at different detection thresholds; the Joint Radar Detection and Communication Coverage Probability (JRDCCP) decreases with the outage threshold and detection threshold, respectively.
Reconfigurable Backscattering Communication System Based on Time Modulation Technique
NI Gang, CHEN Ruihua, HE Chong, JIN Ronghong
2024, 46(6): 2443-2451.   doi: 10.11999/JEIT230700
[Abstract](94) [FullText HTML](73) [PDF 2529KB](24)
Abstract:
In recent years, time-modulated array has aroused much attention due to its superior performance on vector control. Based on the time modulation method, a type of reconfigurable backscattering communication system based on time modulation technique is proposed in this paper. In backscattering node of the proposed system, multiple digital modulation symbols are mapped into the harmonic component of the control waveforms. The scattering or absorbing states of the incoming wave from the base station are then controlled by the designed waveforms. After the receiver samples the backscattering signal and extracts the control waveforms, the digital modulation symbols transmitted from the backscattering node can be recovered from the harmonic component with the Fourier transform. Simulation results demonstrate the performance of the harmonic demodulation methods and consistency with the theoretical values. Meanwhile, the reconfigurable backscattering transmitting experiments based on amplitude, phase shift keying and quadrature amplitude modulation demonstrate the feasibility of the proposed system and methods. In comparison, the proposed system has the characteristics of low power consumption, simple structure and reconfigurable digital modulation.
Outage Performance of Relay-assisted Parasitic Backscatter Communication Networks
SONG Xi, HAN Dongsheng
2024, 46(6): 2452-2461.   doi: 10.11999/JEIT231057
[Abstract](113) [FullText HTML](49) [PDF 1957KB](15)
Abstract:
The existing parasitic backscatter communications rely on the direct links between transceivers and do not work when the direct links are blocked or fade deeply. To solve this problem, a relay-assisted parasitic backscatter communication network is proposed, base on which its outage performance is analyzed. Specifically, according to the proposed network, the instantaneous signal-to-noise ratios to decode the primary and secondary systems are given, and then the outage probabilities of primary and secondary systems on the basis of the energy-causality constraint of the secondary user are defined. Under the Rayleigh channel fading model, the expressions for the outage probability of the primary and secondary systems can be obtained by exploiting mathematical theory. Computer simulations validate the accuracy of the derived primary and secondary system outage probabilities, on which the impacts of different system parameters are analyzed.
Hybrid Reconfigurable Intelligent Surface Assisted Integrated Sensing and Communication: Energy Efficient Beamforming Design
CHU Hongyun, YANG Mengyao, HUANG Hang, ZHENG Ling, PAN Xue, XIAO Ge
2024, 46(6): 2462-2469.   doi: 10.11999/JEIT230699
[Abstract](395) [FullText HTML](423) [PDF 3186KB](104)
Abstract:
Energy Efficiency (EE) is an important design metric for 5G+/6G wireless communications, and Reconfigurable Intelligent Surface (RIS) is widely recognized as a potential means to improve EE. Unlike passive RIS, hybrid RIS consists of both active and passive components, which can amplify the signal strength while phase-shifting the incoming wave, and can effectively overcome the “multiplicative fading” effect caused by fully passive RIS. In view of this, a hybrid RIS-assisted Integrated Sensing and Communication (ISAC) downlink transmission system is proposed in this paper. In order to investigate the intrinsic correlation between data transmission capacity and energy consumption, the paper jointly optimizes the beamforming and phase-shifting of hybrid RIS at the Base Station (BS) under the constraints of BS transmit power, beampattern gain, and hybrid RIS power and amplitude with the goal of maximizing the global EE in a multiuser network. To solve this complex fractional programming problem, an algorithm based on Alternating Optimization (AO) is proposed to solve it. To overcome the problem of high algorithm complexity caused by the introduction of auxiliary variables in the AO algorithm, a solution algorithm based on a cascaded deep learning network is proposed using the association of coupled optimization variables. Simulation results show that the proposed hybrid RIS-assisted ISAC scheme outperforms existing schemes in terms of sum rate and EE, and the algorithm converges quickly.
Network Selection Algorithm Based on Hilbert Space Vector Weighting
MAO Zhongyang, WANG Tingting, LU Faping, ZHANG Zhilin, KANG Jiafang
2024, 46(6): 2470-2479.   doi: 10.11999/JEIT230641
[Abstract](74) [FullText HTML](25) [PDF 3945KB](26)
Abstract:
In order to improve the service completion rate of mobile nodes and the efficiency of network resource allocation in maritime heterogeneous wireless network, a network access selection algorithm based on Hilbert space vector assignment is proposed to address the problems of poor matching between existing network selection algorithms and service demands, and low service completion rate in dynamic environment. The algorithm adopts the network-service matching model based on Hilbert space, maps the network characteristics and service requirements to the same space, and measures whether the network meets the service requirements in the same coordinate system; at the same time, it adopts the pre-switching network selection algorithm based on the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method, and introduces the network-service matching weights to correct the normalization matrix of the distance-to-preferred-solution method, so as to ensure that the selected network matches the service requirements, and to ensure that the network matches the service requirements. This ensures that the selected network matches the service requirements and overcomes the problems of traditional network selection where the service requirements are less considered and the network characteristics and service requirements are difficult to be measured uniformly. In addition, the network switching control algorithm based on spatial distance is adopted, and matching weight and spatial distance are introduced into the network switching control to ensure the continuity of service transmission and improve the service completion rate in the dynamic environment. Simulation results show that compared with the comparison algorithm, the service completion rate of this algorithm is improved by at least 6.81%, which effectively improves the service transmission capacity and smoothness of the network, and indirectly realizes the effective allocation of network resources.
A Joint Optimization Method for Trajectory and Power of Unmanned Aerial Vehicle assisted Over-the-Air Computation
LI Song, LI Jiaqi, WANG Bowen, CHEN Ruirui, SUN Yanjing, ZHANG Xiaoguang
2024, 46(6): 2480-2487.   doi: 10.11999/JEIT230917
[Abstract](266) [FullText HTML](123) [PDF 1382KB](34)
Abstract:
The Unmanned Aerial Vehicle (UAV)assisted over-the-Air Computation(AirComp) system provides an effective solution for the fast aggregation of large-scale and distributed data. In this paper, a joint trajectory planning and power optimization method through UAV-assisted AirComp system is investigated. As a mobile base station, UAV is used to optimize the mean square error of the aggregated data of the AirComp system by adjusting its trajectory and transmitting power of the ground sensors. Under the limitations of UAV trajectory and sensor power, the UAV flight trajectory, the scaling factor and sensor power are jointly optmized to minimize the time-averaged mean square error. Based on the block coordinate descent and successive convex approximation methods, the joint optimization algorithm of UAV flight trajectory and power is proposed. Simulation results verify the performance of the proposed algorithm.
Performance Analysis of Satellite-Aerial-Terrestrial Multiple Primary Users Cognitive Networks Based on NOMA
LIU Rui, GUO Kefeng, ZHU Shibing, LI Changqing, LI Keying
2024, 46(6): 2488-2496.   doi: 10.11999/JEIT230212
[Abstract](135) [FullText HTML](71) [PDF 2655KB](30)
Abstract:
Due to its unique advantages of strong survivability and seamless coverage, Satellite Communication (SatCom) can make up for the shortcomings of ground communication such as terrain limitations and small coverage, and has become increasingly important in current and future communication systems. In addition, aerial-assisted communication is considered as a valuable research direction due to its flexibility and scalability in satellite ground networks. To overcome the problems of spectrum shortage and low spectrum utilization in Integrated Satellite-Aerial-Terrestrial Network (ISATN), Cognitive Radio (CR) and Non-Orthogonal Multiple Access (NOMA) are used in wireless communication networks to improve spectrum utilization and transmission performance. In this regard, the performance of an NOMA-based Cognitive Integrated Satellite-Aerial-Terrestrial Network(CISATN) with multiple primary users is studied, and accurate expressions for Outage Probability (OP) and ergodic capacity of the primary and secondary networks are derived. Asymptotic expressions for the OP and diversity order of these two networks are provided to obtain further insights. Finally, the correctness of the theoretical derivation is verified through numerical simulation, and the impact of key variables on system indicators is analyzed.
Flexible Multiple Access Technology for Satellite Internet of Things
PANG Mingliang, WANG Chaowei, WU Tong, CHEN Jiabin, HUANG Sai, JIANG Fan, ZHANG Junyi
2024, 46(6): 2497-2505.   doi: 10.11999/JEIT231388
[Abstract](144) [FullText HTML](105) [PDF 3907KB](15)
Abstract:
Access latency and complexity in the satellite Internet of Things (IoT) are significantly reduced by the grant-free uplink random access based on Slotted ALOHA (S-ALOHA). However, with the increase of the number of IoT users, the collision probability of S-ALOHA is markedly increased, thereby impacting the performance of the system. This paper addresses the scenario of massive device uplink access in satellite IoT, focusing on the investigation of power resource control for IoT terminals to achieve maximization of system rate. A flexible multiple access scheme based on S-ALOHA is proposed. In the presence of collisions in the system, transmission is carried out using non-orthogonal multiple access technology, which mitigates the issue of repeated transmission of user information and reduces transmission latency. The sequential decision problem of maximizing system rate under the constraint of terminal power is modeled as a Markov process, and the Advantage Actor-Critic (A2C) method is employed to solve it. The simulation results indicate that the success rate of terminal access in scenarios with a massive number of IoT terminals is effectively ensured by the proposed flexible multiple access technology. Additionally, the resource allocation algorithm based on A2C is shown to outperform traditional resource allocation algorithms.
Full Channel Estimation for IRS-assisted Millimeter-wave Mobile Communication Systems Based on Fixed Point Deep Learning
CHU Hongyun, PAN Xue, HUANG Hang, ZHENG Ling, YANG Mengyao, XIAO Ge
2024, 46(6): 2506-2514.   doi: 10.11999/JEIT230692
[Abstract](173) [FullText HTML](80) [PDF 2629KB](22)
Abstract:
Combining Intelligent Reflective Surface (IRS) with massive MIMO can guarantee and improve the performance of millimeter-wave communication systems. An adaptive full-channel estimation method is proposed for the Base Station (BS)-user direct-connect channel and user-IRS-BS reflective channel mixing scenario. First, auxiliary variables are introduced and atomic paradigms are used to correlate the sparse angle-domain subspaces of the direct-connect and reflective channels; then, the full-channel estimation problem is modeled as a continuous angle-domain sparse matrix reconstruction planning by using atomic paradigm minimization; and finally, a low-complexity problem solving algorithm based on the immovable-point deep learning network is designed. The algorithm can not only overcome the dependence of the nonlinear estimation operator on a priori knowledge in the traditional model-based solution method but also adaptively adjust the complexity of the algorithm according to the changes of the mobile scene. Simulation results show that the proposed algorithm can avoid the error propagation effect caused by the traditional time-division estimation strategy, and has higher estimation accuracy and lower complexity.
Interference Performance Evaluation Method Based on Transfer Learning and Parameter Optimization
SUN Zhiguo, XIAO Shuo, WU Yijie, LI Shiming, WANG Zhenduo
2024, 46(6): 2515-2524.   doi: 10.11999/JEIT230817
[Abstract](162) [FullText HTML](108) [PDF 3454KB](23)
Abstract:
A novel method for evaluating interference performance based on Transfer Learning(TL) and parameter optimization is proposed to address the limitation of single evaluation results obtained using traditional error rate assessment in digital communication systems. This method selects the core parameters of each signal processing module as the training index of machine learning and considers the evaluation results of the Technique for Order Preference by Similarity to the Ideal Solution (TOPSIS) as the classification standard. An SVM (Support Vector Machine) is used to train and evaluate the model. The parameter optimization problem in the SVM is addressed by enhancing the global search capability of Ant Colony Optimization (ACO). Moreover, the issue of missing data in the training samples is solved based on the knowledge transfer properties of TL. The results of the simulation experiments demonstrate that the SVM with access to the source domain dataset increases the model accuracy by 4.2%. Parameter optimization, which sacrifices the initial convergence ability, enhances the proximity to the optimal solution by 4.7%.In addition, it can be employed to evaluate the interference performance of digital communication systems.
Distributed Collaborative Path Planning Algorithm for Multiple Autonomous vehicles Based on Digital Twin
TANG Lun, DAI Jun, CHENG Zhangchao, ZHANG Hongpeng, CHEN Qianbin
2024, 46(6): 2525-2532.   doi: 10.11999/JEIT230678
[Abstract](138) [FullText HTML](185) [PDF 3615KB](38)
Abstract:
Focusing on the problems of difficult cooperation between vehicles, low quality of the model trained by cooperation and poor effect of direct application of the obtained results to physical vehicles in the process of path planning for multiple Autonomous Vehicles (AVs), a distributed collaborative path planning algorithm is proposed for multiple AVs based on Digital Twin (DT). The algorithm is based on the Credibility-Weighted Decentralized Federated Reinforcement Learning (CWDFRL) to realize the path planning of multiple AVs. In this paper, the path planning problem of a single AVs is first modeled as the problem of minimizing the average task completion time under the constraints of driving behavior, which is transformed into Markov Decision Process (MDP) and solved by Deep Deterministic Policy Gradient algorithm (DDPG). Then Federated Learning (FL) is used to ensure the cooperation between vehicles. Aiming at the problem of low quality of global model update in centralized FL, this paper uses a decentralized FL training method based on dynamic node selection of reliability to improve the low quality. Finally, the DT is used to assist the training of the Decentralized Federated Reinforcement Learning (DFRL) model, and the trained model can be quickly deployed directly to the real-world AVs by taking advantage of the twin’s ability of learning from DT environment. The simulation results show that compared with the existing methods, the proposed training framework can obtain a higher reward, effectively improve the utilization of the vehicle’s own speed, and at the same time reduce the average task completion time and collision probability of the vehicle swarm.
Radars and Array Signal Processing
Polarized Beam Online Reconfiguration Technique For Flexible Deformation Antennas
CHEN Zhikun, CUI Jinhe, WANG Wei, CHEN Zhibin, GUO Yunfei
2024, 46(6): 2533-2541.   doi: 10.11999/JEIT240070
[Abstract](131) [FullText HTML](48) [PDF 2542KB](22)
Abstract:
In response to the challenges posed by the deformation of flexible polarized array antennas, which results in difficulties in beam reconstruction and compromised beam performance, polarization beam online reconstruction technique based on flexible deformation polarized antennas is proposed in this paper. Firstly, the deformation state of the array is modeled based on a wing model, and real-time deformation data is obtained using modal analysis to reconstruct the antenna array model online. Secondly, the element response of vector array antenna is utilized to construct a flexible array antenna signal model in three-dimensional space. Finally, a deep integration of the Cyclic Algorithm (CA) and Second-Order Cone Programming (SOCP) is employed to solve the dynamic optimization problem of this optimal polarization beam reconstruction. Simulation results demonstrate that within a certain range of deformation and under different arc and angle requirements from environmental loads, the proposed method can achieve online antenna array reconstruction and real-time optimal polarization beam reconstruction based on the dynamic antenna array model. The directional gain, beamwidth, and polarization matching design all meet the requirements for practical engineering applications.
A Hybrid Beamforming Algorithm Based on Limited-Broyden-Fletcher-Goldfarb-Shanno
YAN Junrong, JIANG Peilian, LI Pei
2024, 46(6): 2542-2548.   doi: 10.11999/JEIT230656
[Abstract](126) [FullText HTML](86) [PDF 2017KB](19)
Abstract:
To solve the problems of long runtime, low spectral rate and high bit error rate, which exist in conventional hybrid beamforming schemes, a hybrid beamforming algorithm based on Limited-Broyden-Fletcher-Goldfarb-Shanno (LBFGS) is proposed. Firstly, a single variable objective function is constructed through the least squares solution of the digital precoder. Then, the gradient of the objective function is used to approximate the inverse of the Hessian matrix for obtaining the search direction and the analog precoder is updated along the search direction until the stop condition is satisfied. Finally, the analog precoder is fixed to obtain the digital precoder. The MATLAB simulation analysis indicate that LBFGS algorithm reduces the running time by 28%, increases spectral rate by 1.05%, and reduces bit error rate by 1.06%, compared to MO algorithm.
Electromagnetic Algorithm for Efficiently Analyzing Large Scale Antenna Arrays with Radomes
YIN Lei, HOU Peng, DING Ning, LIN Zhongchao, ZHAO Xunwang, ZHANG Yu, JIAO Yongchang
2024, 46(6): 2549-2557.   doi: 10.11999/JEIT230721
[Abstract](191) [FullText HTML](126) [PDF 7561KB](35)
Abstract:
For the analysis problem of large antenna arrays with radomes, based on the equivalence principle and mode matching theory, the wave port model of Multilevel Fast Multipole Algorithm (MLFMA) is established, and the accurate electromagnetic modeling of antenna excitation source and matching load is realized. Moreover, a parallel strategy of MLFMA for calculating metal-dielectric antenna models is proposed. By establishing multiple octree structures, the communication in processes during the calculation is reduced, and the accurate and efficient analysis of large antenna-array-and-radome-integration system is realized. A comparison of the antenna pattern and S parameters calculated by the proposed algorithm, the higher order method of moments and the finite element-boundary integral is given, validating accuracy and efficiency of the proposed method.
Research on Symmetrically Resonant VLF Transmit/Receive Magnetoelectric Antenna Coupling Performance
WANG Xiaoyu, ZHANG Boyan, ZHAO Xiangchen, YANG Xijie, FENG Qing, CAO Zhenxin
2024, 46(6): 2558-2567.   doi: 10.11999/JEIT230247
[Abstract](99) [FullText HTML](50) [PDF 6406KB](12)
Abstract:
Very low frequency has great potential for long distance signal transmission and military communications due to its low propagation loss characteristics. Magneto-Electric (ME) antennas, based on the acoustic resonance principle, can push the limits of size and are easily impedance matched, offering unique advantages for transmission in the very low frequency band. Based on this, a new ME antenna system consisting of a transmitting antenna of P/T/P structure and a receiving antenna of T/P/T structure is designed. The structural pattern of the antenna in receiving/transmitting electromagnetic waves is analyzed based on the magneto-mechanical coupling model. The magnetic field distribution of the antenna in the near-field range is investigated based on the radiation model. An experiment on the transmission/receiving of ME antennas in the very low frequency band is realized with acoustic wave mediated excitation. Experimental results indicate that obtained at resonant frequencies, the ME transmit/receive antenna is improved by 82.6% in output voltage and by 42.2% in communication range before the structure optimization compared to after the optimization when the piezoelectricity ratio is 0.66 and 0.34, respectively. Magnitude higher radiation efficiency is improved by three orders compared to the same size electric small antenna. Modulated communication with a transmission rate of 5 bit/s is possible and the performance of the antenna is improved based on structural optimization.
Image and Intelligent Information Processing
Efficient File Random Access Method Based on Primer Index Matrix in DNA Storage Scenarios
ZHANG Shufang, LI Yuhui, LI Bingzhi
2024, 46(6): 2568-2577.   doi: 10.11999/JEIT230717
[Abstract](94) [FullText HTML](32) [PDF 3424KB](3)
Abstract:
DNA molecules have the advantages of high density and stability, and are expected to become the medium for the next generation of massive data storage needs, which has received widespread attention in recent years. Currently, primers are used as the unique identifier for files, and random retrieval of DNA pool storage files can be achieved based on polymerase chain reaction (PCR) amplification technology. However, the allocation and mapping relationship between primers and files have not been thoroughly studied, and random allocation is still used to associate primers and files. This will lead to a decrease in the search efficiency of the target primer sequence, and saving the mapping relationship table between primers and files will cause a lot of data redundancy. In order to provide an efficient connection bridge between silicon-based computing devices and carbon-based storage systems, and effectively reduce the data redundancy caused by storing primer-file mapping relationships, a random retrieval method for DNA storage based on the primer index matrix is proposed in this paper. This method constructs a primer index matrix by dividing the stored file set according to different attributes of the file, and converts the primers in the primer library into an ordered primer library according to conversion rules. Finally, the mapping relationship between primers and files is optimized to achieve efficient and multi-dimensional retrieval during file random retrieval. The experimental results show that when storing file sets of different sizes, the efficiency of primer retrieval is improved to a constant level of time complexity by establishing the corresponding primer index matrix using the proposed algorithm in this paper, and the extra storage space required to store the mapping relationship between primers and files is optimized from linear growth to logarithmic growth.
Multi-Scale Occluded Person Re-Identification Guided by Key Fine-Grained Information
ZHOU Yu, ZHAO Xiaofeng, WANG Yi, SUN Yanjing, LI Song
2024, 46(6): 2578-2586.   doi: 10.11999/JEIT230686
[Abstract](254) [FullText HTML](83) [PDF 3160KB](60)
Abstract:
To reduce the influence of background and occlusion on the accuracy of pedestrian identity Re-IDentification (ReID) and make full use of the complementarity between fine-grained and coarse-grained information, a multi-scale occluded pedestrian ReID network guided by key fine-grained information is proposed. First, the image is divided into two types of overlapping patches with different sizes to better simulate the multi-scale characteristics of human observing images and the continuity characteristics of human observing adjacent regions, so a multi-scale recognition network containing both fine-grained and coarse-grained information extraction branches is constructed. Then, considering fine-grained information contains more details and there are similarities and differences between fine-grained and coarse-grained information, fine-grained attention module is further employed to realize the guide of the fine-grained branch to the coarse-grained branch. Among them, the fine-grained information is the key information retained after filtering out the interference information by the Interference Information Elimination (IIE) module. Finally, the key information related to pedestrian ReID is obtained by bivariate difference, and the prediction of pedestrian identity is realized by multi-dimensional joint supervision such as tags and features. Extensive experiments on several public pedestrian ReID databases prove the superiority of this algorithm and the effectiveness and necessity of each module.
Action Recognition Network Combining Spatio-Temporal Adaptive Graph Convolution and Transformer
HAN Zongwang, YANG Han, WU Shiqing, CHEN Long
2024, 46(6): 2587-2595.   doi: 10.11999/JEIT230551
[Abstract](139) [FullText HTML](47) [PDF 4062KB](27)
Abstract:
In a human-centered smart factory, perceiving and understanding workers’ behavior is crucial, as different job categories are often associated with work time and tasks. In this paper, the accuracy of the model’s recognition is improved by combining two approaches, namely adaptive graphs and Transformers, to focus more on the spatiotemporal information of the skeletal structure. Firstly, an adaptive graph method is employed to capture the connectivity relationships beyond the human body skeleton. Furthermore, the Transformer framework is utilized to capture the dynamic temporal variations of the worker’s skeleton. To evaluate the model’s performance, six typical worker action datasets are created for intelligent production line assembly tasks and validated. The results indicate that the model proposed in this article has a Top-1 accuracy comparable to mainstream action recognition models. Finally, the proposed model is compared with several mainstream methods on the publicly available NTU-RGBD and Skeleton-Kinetics datasets, and the experimental results demonstrate the robustness of the model proposed in this paper.
Spiking Sequence Label-Based Spatio-Temporal Back-Propagation Algorithm for Training Deep Spiking Neural Networks
WANG Zihua, YE Ying, LIU Hongyun, XU Yan, FAN Yubo, WANG Weidong
2024, 46(6): 2596-2604.   doi: 10.11999/JEIT230705
[Abstract](136) [FullText HTML](71) [PDF 1193KB](22)
Abstract:
Spiking Neural Networks (SNN) have a signal processing mode close to the cerebral cortex, which is considered to be an important approach to realize brain-inspired computing. However, the lack of effective supervised learning algorithms for deep spiking neural networks has been a great challenge for spiking sequence label-based brain-inspired computing tasks. A supervised learning algorithm for training deep spiking neural network is proposed in this paper. It is an error backpropagation algorithm that uses surrogate gradient to solve the problem of non-differentiable spike generation function, and define the postsynaptic potential and membrane potential reversal iteration factors represent the spatial and temporal dependencies of pulsed neurons, respectively. It differs from existing learning algorithms based on firing rate encoding, fully reflects analytically the temporal dynamic properties of the spiking neuron. Therefore, the algorithm proposed in this paper is well-suited for application to tasks that require longer time sequences rather than spiking firing rates, such as behavior control. The proposed algorithm is validated on the static image datasets CIFAR10, and the neuromorphic dataset NMNIST. It shows good performance on all these datasets, which helps to further investigate spike-based brain-inspired computation.
Multi-Scale Attention Recurrent Network with Multi-order Taylor Differential Knowledge for Deep Spatiotemporal Sequence Prediction
SUN Qiang, ZHAO Ke
2024, 46(6): 2605-2618.   doi: 10.11999/JEIT231108
[Abstract](175) [FullText HTML](47) [PDF 10791KB](52)
Abstract:
Deep spatiotemporal sequence prediction methods that incorporate a priori physical knowledge are commonly characterized by the utilization of Partial Differential Equations (PDE) for modeling. However, two main issues are concerned: (1) the limited precision in approximations with PDEs; and (2) the inability to efficiently capture spatiotemporal features at multiple spatial scales as well as the edge spatial information of the spatiotemporal sequences in the recurrent network. To address these challenges, one Taylor Differential Incorporated Convolutional Recurrent Neural Network (TDI-CRNN) is proposed in this paper. Firstly, in order to enhance the approximation accuracy of higher-order partial differential equations and to alleviate the limitations of PDE applications, one physical module with multi-order Taylor approximation is designed. The module is firstly used for the differential approximation of the input sequence by means of the Taylor expansion, and then couples the differential convolution layers with different orders via differential coefficients, and dynamically adjusts the truncation order and the number of differential terms of the Taylor expansions. Secondly, to capture the multiple spatial scale features of the hidden states in the recurrent network and to better capture the edge spatial information of the spatiotemporal sequences, one Multi-Scale Attention Recurrent Module (MSARM) is devised. Multi-scale convolution and spatial attention mechanisms are utilized in the convolution layer of the Multi-scale Convolution Spatial Attention UNet (MCSA-UNet), aiming to focus on local spatial regions within spatiotemporal sequences. Extensive experiments are conducted on the Moving MNIST, KTH, and CIKM datasets. The Mean Squared Error (MSE) on the Moving MNIST dataset dropped to 42.7, while the Structural Similarity Index Measure (SSIM) increased to 0.912. The SSIM and Peak Signal-to-Noise Ratio (PSNR) on the KTH dataset increased to 0.882 and 29.03, respectively. The Correct Skill Index (CSI) on the real weather radar echo CIKM dataset increased to 0.515. The final visualization and quantitative prediction results verify the rationality and effectiveness of the TDI-CRNN model.
Network and Information Security
A Computational Offloading Incentive Forward Contract Taking into Account Risk Appetite
ZHANG Biling, JIAO Zhengyang, LIU Jiahua, GUO Caili
2024, 46(6): 2619-2626.   doi: 10.11999/JEIT230617
[Abstract](109) [FullText HTML](32) [PDF 2843KB](14)
Abstract:
In edge computing networks, to stimulate the Edge Computing Nodes (ECNs) to assist in computation offloading to relieve the pressure of computing Service Provider (SP), a forward transaction oriented incentive mechanism is studied. Considering that there is information asymmetry between SP and ECNs and the uncertainty of ECNs idle resources can lead to cooperation risks, a risk-aware forward incentive mechanism based on contract theory for computation offloading is proposed. Firstly, a risk preference model for nodes is established; and then the Individual Rationality (IR) constraints and Incentive Compatibility (IC) constraints are defined, and the incentive problem is modeled as a forward contract design problem to maximize the benefits of SP; finally, the optimal forward contract is derived after constraint simplification. The simulation results verify the feasibility and rationality of the proposed forward contract, and prove that the contract can effectively incentivize ECNs to participate in computation offloading and increase the profits of SP.
Intelligent Semantic Location Privacy Protection Method for Location Based Services in Three-Dimensional Spaces
MIN Minghui, YANG Shuang, XU Junhuai, LI Xin, LI Shiyin, XIAO Liang, PENG Guojun
2024, 46(6): 2627-2637.   doi: 10.11999/JEIT230708
[Abstract](148) [FullText HTML](50) [PDF 4589KB](21)
Abstract:
An intelligent semantic location privacy protection method based on 3D Geo-Indistinguishability (3D-GI) is studied for the privacy leakage problem of sensitive semantic locations (such as medicine stores and bookstores) in 3D space location-based services, such as hospitals and shopping centers. Reinforcement Learning (RL) techniques are used in this paper to optimize user’s semantic location privacy protection policies dynamically. Specifically, a 3D semantic location perturbation mechanism is proposed based on the Policy Hill Climbing (PHC) algorithm, independent of specific environments and attack models. This mechanism induces attackers to infer less sensitive locations to reduce the exposure of sensitive semantic locations. To address the dimensional disaster problem of complex 3D space, a 3D semantic location perturbation mechanism based on the Proximal Policy Optimization (PPO) algorithm is further proposed. This mechanism captures the environment features using a neural network and optimizes the neural network parameter updates through the offline policy gradient method to improve the efficiency of semantic location perturbation policy selection. Experimental results show that the proposed mechanism improves both semantic location privacy protection and user service experience.
A Cloud Server Anomaly Detection Model Based on Time Series Decomposition and Spatiotemporal Information Extraction
TANG Lun, ZHAO Yuchen, XUE Chengcheng, CHEN Qianbin
2024, 46(6): 2638-2646.   doi: 10.11999/JEIT230679
[Abstract](168) [FullText HTML](65) [PDF 4836KB](37)
Abstract:
Anomaly detection is an important task to maintain cloud data center performance. A large number of cloud servers are running in cloud data centers to implement various cloud computing functions. Since the performance of cloud data centers depends on the normal operation of cloud services, it is crucial to detect and analyze anomalies in cloud servers. To this end, a cloud server anomaly detection model based on time series decomposition and spatiotemporal information extraction-Multi-Channel Bidirectional Wasserstein Generative Adversarial Networks with Graph-Time Network (MCBiWGAN-GTN) is proposed in this paper. Firstly, the Bidirectional Wasserstein GAN with Graph-Time Network (BiWGAN-GTN) algorithm is proposed. This algorithm is built upon the Bidirectional Wasserstein GAN with Gradient Penalty (BiWGAN-GP) algorithm. In this modification, the generator and encoder are replaced by a spatiotemporal information extraction module— Graph-Time Network (GTN) composed of Graph Convolutional Networks (GCN) and Temporal Convolutional Networks (TCN). This modification aims to extract spatiotemporal information from the data, enhancing the capabilities of the algorithm. Secondly, the semi-supervised BiWGAN-GTN algorithm is proposed to identify anomalies in multi-dimensional time series to avoid the risk of abnormal data intrusion during the training process and enhance model robustness. Finally, the MCBiWGAN-GTN is designed to achieve the goal of reducing data complexity and improving model learning efficiency. The Complete Ensemble Empirical Mode Decomposition with Adaptive Noise algorithm (CEEMDAN) is used to decompose the time series data, and then different components are sent to the BiWGAN-GTN algorithm under the corresponding channel for training. The effectiveness and stability of the proposed model are verified on two real-world cloud data center datasets, Clearwater and MBD, using three evaluation metrics: precision, recall and F1 score. Experimental results show that the performance of MCBiWGAN-GTN on these two datasets is stable and better than the compared methods.
Abnormal Traffic Detection Method Based on Traffic Spatial-temporal Features and Adaptive Weighting Coefficients
GU Wei, XING Hongyan, HOU Tianhao
2024, 46(6): 2647-2654.   doi: 10.11999/JEIT230825
[Abstract](360) [FullText HTML](108) [PDF 2205KB](74)
Abstract:
Considering the problem that the performance of the traditional abnormal traffic detection models is limited by the low utilization of spatiotemporal features of traffic data, an abnormal traffic detection method MSECNN-BiLSTM based on the combination of Convolutional Neural Network (CNN), Multi head Squeeze Excitation mechanism (MSE), and Bidirectional Long Short-Term Memory (BiLSTM) network is proposed. The one-dimensional CNN is used to capture abnormal traffic features at spatial scales. The MSE mechanism is introduced to adaptively calibrate the feature weights and strengthen the model’s ability to correlate global features from multiple perspectives. The traffic features are input into BiLSTM to capture the temporal dependencies of the traffic data and further model the relationship of network traffic on the time scale. The softmax classifier is employed for traffic detection. The experimental results verify that the proposed model is effective in the field of abnormal traffic detection.
Improved Meet-in-the-middle Attacks on Reduced-round E2
DU Xiaoni, SUN Rui, ZHENG Yanan, LIANG Lifang
2024, 46(6): 2655-2662.   doi: 10.11999/JEIT230655
[Abstract](162) [FullText HTML](55) [PDF 2732KB](44)
Abstract:
E2 is one of the 15 candidate algorithms in the first round of AES, which has the characteristics of excellent software and hardware implementation efficiency and strong security. The meet-in-the-middle attacks on E2 are carried out in this paper by using multiset tabulation technique and differential enumeration technique. First, E2-128 is taken as an example to improve the existing 4-round meet-in-the-middle distinguisher, and the pre-computation complexity of 5-round key recovery attack is reduced to \begin{document}${2^{31}}$\end{document} 5-round encryptions. Second, for E2-256, a 6-round distinguisher is constructed from the new 4-round distinguisher by extending two rounds backward, and then a 9-round meet-in-the-middle attack is presented, whose data complexity is \begin{document}${2^{105}}$\end{document} chosen plaintexts, memory complexity is \begin{document}${2^{200}}$\end{document} Byte, and time complexity is \begin{document}${2^{205}}$\end{document} 9-round encryptions. Compared with the existing security analysis results of E2, the scheme achieves the longest number of attack rounds for E2-256.
Circuit and System Design
FPGA-Based Unified Accelerator for Convolutional Neural Network and Vision Transformer
LI Tianyang, ZHANG Fan, WANG Song, CAO Wei, CHEN Li
2024, 46(6): 2663-2672.   doi: 10.11999/JEIT230713
[Abstract](897) [FullText HTML](455) [PDF 3611KB](230)
Abstract:
Considering the problem that traditional Field Programmable Gate Array (FPGA)-based Convolutional Neural Network(CNN) accelerators in computer vision are not adapted to Vision Transformer networks, a unified FPGA accelerator for convolutional neural networks and Transformer is proposed. First, a generalized computation mapping method for FPGA is proposed based on the characteristics of convolution and attention mechanisms. Second, a nonlinear and normalized acceleration unit is proposed to provide acceleration support for multiple nonlinear operations in computer vision networks. Then, we implemented the accelerator design on Xilinx XCVU37P FPGA. Experimental results show that the proposed nonlinear acceleration unit improves the throughput while causing only a small accuracy loss. ResNet-50 and ViT-B/16 achieved 589.94 GOPS and 564.76 GOPS performance on the proposed FPGA accelerator. Compared to the GPU implementation, energy efficiency is improved by a factor of 5.19 and 7.17, respectively. Compared with other large FPGA-based designs, the energy efficiency is significantly improved, and the computing efficiency is increased by 8.02%~177.53% compared to other FPGA accelerators.
In-memory Wallace Tree Multipliers Based on Majority Gates with Voltage Gated Spin-Orbit Torque Magnetoresistive Random Access Memory Devices
HUI Yajuan, LI Qingzhen, WANG Leimin, LIU Cheng
2024, 46(6): 2673-2680.   doi: 10.11999/JEIT230815
[Abstract](352) [FullText HTML](165) [PDF 5111KB](56)
Abstract:
In the research on utilizing emerging non-volatile storage arrays for in-memory computing, the latency of in-memory multipliers often exhibits exponential growth with increasing bit width, and significantly impacts the computational performance. A Voltage-Gated Spin-Orbit Torque Magnetoresistive Random-Acess Memory (VGSOT-MRAM) device unit crossbar array is proposed and a circuit design approach for in-memory Wallace tree multipliers is presented in this paper. The proposed series-connected storage unit structure effectively addresses the issue of low resistance values in magnetic storage units through resistive summing. Furthermore, an in-memory computing architecture based on a voltage-controlled spin-orbit torque magnetic storage unit crossbar array is introduced. Finally, a five-input majority decision logic gate implemented during the “read” operation is leveraged to further reduce the logic depth of the Wallace tree multiplier. Compared to existing multiplier design methods, the proposed approach reduces the delay overhead from O(n2) to O(log2 n), with even lower latency for larger bit widths.
News
more >
Conference
more >
Author Center

Wechat Community