Current Issue

• • 2022, Vol. 45 No. 2 Published date:28 April 2022 Last issue Next issue

  REVIEW

  • Adaptive Integrated Navigation Based Artificial Intelligence
  • ZHAO Fang, WU Fan
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 1-8. DOI:10.13190/j.jbupt.2021-157
  • Abstract ( 1076 )     HTML( 857 )     　　
  • This work first introduces and classifies the adaptive integrated navigation algorithms in global navigation satellite systems and inertial navigation systems. According to the characteristics of the adaptive integrated navigation algorithm in different directions, this work further subdivides and classifies each kind of adaptive integrated navigation algorithm, and concretely introduces the representative algorithm in each class. The advantages and disadvantages of each representative algorithm are analyzed. Finally, this work summarizes the difficulties and challenges of the adaptive integrated navigation, and shed some light on the future research trend in this field.
  • References | Supplementary Material | Related Articles

  PAPERS

  • Service Function Chain Embedding for Network Slicing with Diversified Requirements
  • LI Hang, WEN Xiangming, KONG Zixuan, XIANG Wan, WANG Luhan
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 9-15. DOI:10.13190/j.jbupt.2021-165
  • Abstract ( 492 )     HTML( 183 )     　　
  • Network slicing realizes the creation and orchestration of slices through service function chain embedding. However, the service function chain embedding in network slicing should consider diversified requirements of service function chains, reduce deployment costs and increase the acceptance ratio by taking virtual network function sharing and admission control into consideration. The service chain embedding for network slicing with diversified requirements is studied, where virtual network function sharing and admission control are introduced. First, the above problem is formulated as an optimization model that maximizes the network net revenue, and then an embedding algorithm is devised. The simulation results show that the proposed algorithm outperforms the existing benchmark algorithm, and achieves near-optimal performance.
  • References | Supplementary Material | Related Articles

  • A Risk Assessment Method of Network Host Node with Host Importance
  • YANG Hongyu, YUAN Haihang, ZHANG Liang
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 16-21. DOI:10.13190/j.jbupt.2021-166
  • Abstract ( 363 )     HTML( 301 )     　　
  • The existing network host node assessment methods based on attack graph have unreasonable calculation of atomic attack probability and asset protection value, and ignore the impact of the association relationship between host nodes on host node risk value. To solve these problems, a risk assessment scheme based on the importance of hosts in the network is proposed. Firstly, a host-based attack graph based on network information is build, and then the probability of atomic attack is calculated by vulnerability exploitability, code availability and defense intensity. After that, the attack probability of path is calculated based on the attack graph. Furthermore, the host importance is characterized from the attack graph structure and asset protection value. The reciprocal of atomic attack probability is used to weigh the host-based attack graph, and the improved weighted betweenness index is calculated. Moreover, the entropy weight method is used to weigh the asset protection value index of the host nodes and calculate the asset protection value. Finally, the risk value of the network host node is calculated according to the maximum path attack probability and host importance. The experimental results show that the host node risk in the network environmentand the obtained risk value results are more rational.
  • References | Supplementary Material | Related Articles

  • Hardware Model-Aware Joint Offloading and Resources Allocation Optimization
  • ZHI Jialin, WANG Nan, MAN Yi, TENG Yinglei
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 22-28. DOI:10.13190/j.jbupt.2021-143
  • Abstract ( 499 )     HTML( 315 )     　　
  • The traditional researches on edge computing offloading do not involve the details of computer hardware implementation. In addition, their computing model is rough, and the optimization scheme has low accuracy. To solve these problems, a hardware-based scheme that jointly optimizes computing offloading and resource allocation for multi-user multi-edge server is proposed. In particular, the details of hardware implementation in the calculation process are considered. From the perspective of the granularity of computer instruction execution, the input/output bottleneck and the energy consumption of memory functional modules are calculated first. Then, the joint optimization model is re-established. Finally, the system energy consumption is minimized on the premise of meeting the delay requirement of offloading tasks. To solve the high-dimensional problem of action space, a hybrid online two-part matching kuhn-munkras algorithm based on deep deterministic policy gradient is adopted. The simulation results show that the memory energy consumption cannot be ignored. Besides, the proposed optimization algorithm can effectively learn the optimal strategy and has a significant effect on reducing the system energy consumption.
  • References | Supplementary Material | Related Articles

  • Adaptive Kernel RBFNN Based on Normalized Least Mean Square Algorithm
  • HUO Yuanlian, GONG Qi, QI Yongfeng, AN Yaqi
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 29-35. DOI:10.13190/j.jbupt.2021-132
  • Abstract ( 499 )     HTML( 237 )     　　
  • To make the adaptive kernel radial basis function neural network (RBFNN) exhibit the characteristics of fast convergence and steady-state error, a method that optimizes the adaptive kernel RBFNN by using the normalized least mean square as the learning algorithm is proposed. Based on the gradient descent algorithm, we derive the normalized least mean square (NLMS) algorithm with a variable step factor, and use it as a learning algorithm to update the weights and the biases of the adaptive kernel RBFNN. The simulation results in nonlinear system identification and pattern classification show that using NLMS learning algorithm to train adaptive kernel RBFNN has faster convergence speed and relatively less steady-state error compared with other learning algorithms.
  • References | Supplementary Material | Related Articles

  • EEG-Based Emotion Recognition by Using Convolutional Echo-State Network
  • CHAO Hao, MA Qingmin, LIU Yongli
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 36-43. DOI:10.13190/j.jbupt.2021-078
  • Abstract ( 461 )     HTML( 213 )     　　
  • A convolutional echo-state network (CESN) model is proposed for the emotion recognition task based on electroencephalogram (EEG) signals. First, a feature matrix sequence of EEG signal is constructed. Then, high-level abstract features are extracted from the feature matrices via convolution, and one dimension feature vectors are formed. After that, a reservoir with self-feedback function is employed to extract the dynamic temporal information from the feature vector sequence. Finally, the emotion recognition task is realized by ridge regression. The experiment is carried out on the database for emotion analysis using physiological signals datasets. The experiment result shows that the EEG signal segments contain temporal information related to emotion, and the CESN model can mine and utilize the information effectively. In addition, the proposed CESN model can circumvent the problems of the local optimization and long training time, which are caused by back propagation algorithm in convolutional neural network.
  • References | Supplementary Material | Related Articles

  • Satellite Clock Bias Prediction Based on GM(1,1) and D-MECM
  • CHENG Jiahui, MIAO Xinyu, ZHAO Jingyan, QIAO Yaojun, YU Song
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 44-49. DOI:10.13190/j.jbupt.2021-149
  • Abstract ( 341 )     HTML( 237 )     　　
  • To improve the accuracy in medium-term and long-term prediction of a single satellite clock bias prediction model when the amount of modeling data is small, a combined prediction model based on grey model and first-order difference modified exponential curve method is proposed. In the model, a small amount of historical clock data is used to build a grey model and predict the clock data in the future, and then the prediction data is used as the modeling data of the first order difference modified exponential curve model to make the medium-term and long-term prediction of the clock. The simulation results show that the combined prediction model can forecast the clock difference with high accuracy based on a small amount of historical data. The experiment is carried out based on the precise clock difference data collected by the satellite common view instrument. Compared with the single quadratic polynomial model and the gray model, the results show that when using 5h clock difference data for modeling and forecasting 48h clock difference data in the future, the average prediction accuracy of quadratic polynomial model and gray model is 285.06ns and 91.11ns, respectively, while the average prediction accuracy of combined model is 29.48ns. Thus, compared with the single quadratic polynomial model and gray model, the proposed combined prediction model achieves 89.66% and 67.64% accuracies gains, respectively.
  • References | Supplementary Material | Related Articles

  • WSN Routing Algorithm Based on Firefly Algorithm to Optimize FCM
  • YU Xiuwu, QIN Xiaokun, LIU Yong
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 50-56. DOI:10.13190/j.jbupt.2021-154
  • Abstract ( 418 )     HTML( 227 )     　　
  • The limited energy of nodes in wireless sensor networks is prone to unbalanced energy load. To solve this problem, a routing algorithm for wireless sensor networks (WSN) is proposed based on the firefly algorithm to optimize fuzzy C-means (FCM). Both the clustering stage and the inter-cluster routing establishment stage in clustering routing algorithm are optimized in firefly algorithm to optimize fuzzy C-means (FFACM) algorithm. In the clustering stage, the firefly algorithm is used to calculate the initial clustering center, so as to avoid the problem that fuzzy C-means algorithm falls into local optimum due to the initial clustering center. To select the cluster head node, the fitness function about residual energy and distance is established, and the cluster head node with the largest fitness value is selected and dynamically updated. We calculate the link cost between nodes and establish the cost function according to the residual energy and the distance to the sink node, and then select the node with the smallest cost function value to establish the multi-hop routing between clusters, which minimizes the load of the cluster head node. The simulation results show that compared with other routing algorithms in wireless sensor networks, FFACM algorithm can effectively balance the network load, reduce the energy consumption of nodes, and thus it can prolong the network life cycle.
  • References | Supplementary Material | Related Articles

  • Incremental Dehazing Algorithm Combining Feature Enhancement and Multi-Scale Loss
  • WANG Keping, WEI Jinyang, YANG Yi, FEI Shumin, CUI Kefei
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 57-64. DOI:10.13190/j.jbupt.2021-106
  • Abstract ( 414 )     HTML( 205 )     　　
  • To improve the clarity of the haze image and realize the generalization of the dehazing ability, a network structure with feature enhancement and multi-scale loss constraint is proposed, which is trained by an incremental training method. The network consists of a teacher network and a student network. The student network enhances the features by learning the attention information of the labeled samples extracted by the teacher network and uses the multi-scale semantic features of the labeled samples as the soft targets. Besides, a multi-scale semantic feature loss measurement mechanism is established, which cascades with the global pixel difference loss to construct the loss function for feature and pixel levels. According to the incremental training method, the teacher network guides the student network to balance the relationship between the new and old knowledge of different datasets. Thus, the network can improve the generalization ability of the supplementary dataset quickly, while retains the original knowledge. Experiment results show that the proposed algorithm performs well in both subjective visual effects and objective evaluation indicators.
  • References | Supplementary Material | Related Articles

  • A Joint Intelligent Optimization Scheme of Computation Offloading and Resource Allocation for MEC
  • DU Mei, ZHOU Junhua, LI Dunqiao, CHEN Shizhao, WEI Yifei
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 65-71. DOI:10.13190/j.jbupt.2021-145
  • Abstract ( 945 )     HTML( 412 )     　　
  • Due to the distributed base station deployment, limited server resources and dynamic end-users in mobile edge computing (MEC), the design of computation offloading scheme is extremely challenging. Since the deep reinforcement learning has advantages in terms of dealing with dynamic complex problems, we design the optimal computation offloading and resource allocation strategies based on deep reinforcement learning to minimize the system energy consumption. First, the network framework of cloud edge-end collaboration is considered. Then, the joint computation offloading and resource allocation problem is defined as a Markov decision process.Next, a multi-agent deep deterministic policy gradient-based learning algorithm is proposed to minimize the system energy consumption. The experimental results show that our proposed scheme significantly reduces the energy consumption compared to the deep deterministic policy gradient-based algorithm and the full offloading policy.
  • References | Supplementary Material | Related Articles

  • A Road-Level Traffic Accident Risk Prediction Method
  • NING Jing, SHE Hongyan, ZHAO Dong, LUO Dan, WANG Lei
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 72-78. DOI:10.13190/j.jbupt.2021-142
  • Abstract ( 457 )     HTML( 519 )     　　
  • Existing deep-learning-based methods always divide the predicted region into grids which does not conform to the natural form of accidents,while accidents generally occur on roads. Aiming at the problem of road-level accident risk prediction, an urban traffic accident risk prediction model scale-reduced attention based on graph convolution network (SA-GCN) is proposed. First, the model effectively combines historical long-term and short-term risks, external weather features and a gated graph convolution structure to capture spatial-temporal correlations, and then an attention mechanism is applied to obtain dynamic representations of spatial-temporal features. After that, to solve the problem of the sparseness and spatial heterogeneity of accident data, a scale reduction module which uses the accident risk of the coarse-grained area after clustering, is designed to guide the accident risk prediction at road level. Experimental results on real traffic datasets performance measurement system show that the SA-GCN model performs better than six baseline models, and achieves 11% higher prediction accuracy than the state-of-the-art model.
  • References | Supplementary Material | Related Articles

  • Driver Distraction Recognition Using Bilinear Fusion Networks
  • LIU Changyuan, HU Haoyuan, BI Xiaojun
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 79-84. DOI:10.13190/j.jbupt.2021-171
  • Abstract ( 425 )     HTML( 175 )     　　
  • Accurate recognition of driver's distraction behavior can radically reduce the traffic accidents. Traditional recognition methods have the problems of few classification categories and low accuracy. To solve these problems, residual neural network (ResNet-50) is employed to recognize driver's distraction behavior and improve the network. To further improve the feature extraction ability of the model, bilinear fusion is carried out on the features extracted from the improved ResNet-50 model and EfficientNet-B0 model. Thus, the recognition accuracy of the model can be further improved. The average accuracy of the improved ResNet-50 single model is up to 94.2%, and the average accuracy of the model after fusing the improved ResNet-50 with EfficientNet-B0 is up to 96.7%. Experimental results show that this method has a good classification effect on the detection of driver's distraction behavior.
  • References | Supplementary Material | Related Articles

  • Service Function Chain Orchestrating Algorithm Based on SDN and NFV
  • JIA Yuning, WEI Yifei, ZHOU Junhua
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 85-90. DOI:10.13190/j.jbupt.2021-218
  • Abstract ( 748 )     HTML( 381 )     　　
  • By combining software defined network (SDN) and network function virtualization (NFV), and applying them to the resource orchestration problem of service function chains (SFC) deployment, the three-tier architecture consisting of SDN controller, network function virtualization and physical underlying computing resource layer in the process of heterogeneous network resource mapping is considered. An optimization algorithm for active control resources based on SDN and NFV is proposed. First, the user's utility is modeled by the multi-standard aggregated multi-criteria utility algorithm, and the optimization goal is transformed into a user's utility maximum problem. Then, based on the algorithm's prediction of the future state and real-time monitoring of the network utilization, the controller makes decisions and issues control commands for the arriving SFC requests, which is then used to occupy the underlying resources held by the virtual network function. The simulation results show that compared with the static timing resource allocation algorithm, the proposed active control algorithm performs better in terms of resource utilization, acceptance rate, and user creation utility.
  • References | Supplementary Material | Related Articles

  • Joint Optimization Algorithm for Task Offloading and Resource Allocation in Heterogeneous Networks
  • ZHANG Yuqing, LI Yun, HUANG Hongrui, ZHUANG Hongcheng
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 91-97. DOI:10.13190/j.jbupt.2021-204
  • Abstract ( 535 )     HTML( 298 )     　　
  • Under the constraint of limited network edge resources, considering the impact of service diversity and network access heterogeneity on task offloading and computational resource allocation, a joint optimization algorithm for task offloading and computing power resources allocation under heterogeneous network is proposed to jointly handle local and server tasks. The proposed algorithm makes a trade-off between system revenue and task offloading delay by invoking Lyapunov theory and search tree algorithm. Furthermore, to quickly branch and bound the search tree, the offloading priority criterion is designed. Finally, the simulation results verify the effectiveness and rationality of the proposed algorithm.
  • References | Supplementary Material | Related Articles

  • Channel Characterization of Nonreciprocal Beam Patterns Based on the 3D Geometric Channel Model
  • ZHANG Jiachi, LIU Liu, LI Lu, TAN Zhenhui, ZHOU Tao
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 98-103. DOI:10.13190/j.jbupt.2021-177
  • Abstract ( 359 )     HTML( 266 )     　　
  • To investigate the influence of nonreciprocal beam patterns (NBP) on the wireless channel characteristics, a 3-dimensional (3D) geometric channel model is used to analyze the small-scale fading characteristics of the millimeter-wave wireless channel under this pattern in a macro-cell scenario. Specifically, a non-stationary 3D geometric channel modeling method is proposed based on the propagation graph theory. Moreover, by adopting transceiver beams, effective scatters are found out to reduce the computational complexity. According the proposed model, the influence of NBP on the selective fading characteristics involving time, frequency, and space domains are also analyzed. Simulation results reveal that the number of effective scatters located inside the intersection of downlink transceiver beams is larger than that of the uplink case, and thus the dispersion effects in Doppler, delay, and angular spreads of downlink transceiver beams are more serious than these of the uplink case. Besides, the latter case can be approximated as a line-of-sight scenario with a single propagation path.
  • References | Supplementary Material | Related Articles

  • Clustering Algorithm Combined with Discriminant Function in Ultra Dense Network
  • KANG Ling, WANG Yi, HU Yanjun, JIANG Fang, LI Liping
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 104-109. DOI:10.13190/j.jbupt.2021-180
  • Abstract ( 234 )     HTML( 171 )     　　
  • Ultra-dense networks can enhance user experience through collaboration between virtual cells, while the overlapping coverage of cells makes the interference problem between users more complex. Therefore, a discriminant function-based clustering algorithm is proposed to mitigate the throughput degradation problem caused by strong interference. Firstly, the inter-user interference network is defined based on the cosine similarity of inter-user interference channels. Then, cluster heads are selected and users are classified based on the interference network. Meanwhile, to solve the fuzzy user belonging to clusters under virtual cells, a discriminant function is designed to fuzzy-classify users based on the principle of maximizing the sum of inter-cluster interference weights and minimizing the sum of intra-cluster interference weights. The simulation results show that compared with the existing methods, the proposed algorithm improves the system throughput by 10%-30% without increasing the complexity, and has certain advantages for edge users.
  • References | Supplementary Material | Related Articles

  REPORTS

  • ICNN Fault Diagnosis Method Based on EEMD
  • XU Tongle, MENG Liang, KONG Xiaojia, SU Yuanhao, SUN Yanfei
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 110-116. DOI:10.13190/j.jbupt.2021-128
  • Abstract ( 333 )     HTML( 114 )     　　
  • An improved convolutional neural network fault diagnosis method based on ensemble empirical mode decomposition is proposed to extract weak fault features and low fault diagnosis accuracy of bearings. Firstly, we employ ensemble empirical mode decomposition to reduce the noise of the signal, and transform the denoised signal into a 2-dimensional (2-D) signal. Then, to solve the problem of parameter explosion of convolutional neural network (CNN) and the uncertainty of data characteristics, a batch normalization layer is added between the convolutional layer and the pooling layer of CNN to construct an improved convolutional neural network (ICNN). Finally, the weak fault dataset of the wind turbine bearings are taken as an example to verify that the proposed method has superior performance compared with other methods. The experimental results show that the proposed method can effectively extract fault features, and has high fault diagnosis accuracy and efficiency.
  • References | Supplementary Material | Related Articles

  • A High-Performance Downlink Synchronization Algorithm Based on Convolutional Neural Network for 5G Systems
  • LI Xiaohui, WANG Xianwen, FAN Tao, LIU Jiawen, WAN Hongjie
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 117-123. DOI:10.13190/j.jbupt.2021-086
  • Abstract ( 433 )     HTML( 205 )     　　
  • To solve the problem of low success rate of the fifth generation of mobile communications system (5G) downlink synchronization in low signal-to-noise ratio and large frequency offset environment, a synchronization signal block (SSB) detection algorithm based on convolutional neural network (CNN) and an improved hybrid correlation synchronization algorithm are proposed. Without the prior information, the wireless signal is segmented by the maximum autocorrelation criterion and the characteristics of the cyclic prefix to generate data sets. Then, a CNN is constructed to detect any SSB carried by beams, which can locate the SSB target interval quickly and reduce the search range in the correlation process. Furthermore, the improved hybrid correlation algorithm is used to complete primary synchronization signal timing synchronization and frequency offset estimation in the target interval. Simulation results show that the proposed algorithms preform well in terms of SSB detection rate and timing synchronization, and can resist the influence of noise and frequency offset effectively.
  • References | Supplementary Material | Related Articles

  • Transformer Based Scene Character Detection over Low Quality Images
  • ZHANG Chongsheng, CHEN Jie, ZONG Ruixing, YANG Shuailei, FAN Gaojuan
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 124-130. DOI:10.13190/j.jbupt.2021-155
  • Abstract ( 585 )     HTML( 278 )     　　
  • In order to solve the problem of character-level scene text detection and recognition under imperfect imaging conditions, a Transformer based scene character detection algorithm is proposed. Firstly, a Transformer based encoding-decoding structure is designed which takes the order of characters in the text instances into account, so as to output the position and order of sequence information of each character detection box can be output. Then, the Hungarian algorithm is used to calculate the loss of the algorithm which combines bounding box coordinates and ranking losses. Finally, through the experiments on three character-level annotated data sets, we show that under different evaluation metrics, the proposed method is able to achieve good performance on in terms of both scene character localization and recognition.
  • References | Supplementary Material | Related Articles

  • High Performance SC-LDPC Code Based on Affine Permutation Matrices
  • ZHU Kun, YANG Hongwen
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(2): 131-136. DOI:10.13190/j.jbupt.2021-175
  • Abstract ( 342 )     HTML( 230 )     　　
  • Spatially-coupled (SC) low density parity check (LDPC) code is a kind of LDPC code with convolution structure. This structure not only brings convolution gain but also introduces memory structure. For codes using sliding window decoding, the previous error message will affect the subsequent decoding, especially for SC-LDPC codes with long coupling length, which is easy to cause error propagation. Therefore, SC-LDPC codes have stricter requirements for structural design than traditional LDPC block codes. To improve the design space and performance of SC-LDPC codes, an affine permutation matrix (APM) is proposed to replace the traditional cyclic permutation matrix for code construction. The full cycles and non-full cycles phenomena in APM-LDPC code's structures are discovered and demonstrated, and the non-full cycles phenomenon only occurs in APM-LDPC code is proved. The APM-SC-LDPC code with non-full cycles phenomenon can significantly reduce the number of short cycles, the error-floor and significantly improve in waterfall region.
  • References | Supplementary Material | Related Articles