Current Issue

• • 2022, Vol. 45 No. 4 Published date:28 August 2022 Last issue Next issue

  • Design of Three-dimensional Convolution Neural Network Model for Intelligent Aided Diagnosis of Coarctation of Aorta
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 1-6.
  • Abstract ( 155 )     　　
  • Combining the 3D spatial features of cardiac CT images, a model of coarctation of aorta (CoA) based on 3D convolutional neural network (C3D) is proposed (3DCoA). Compared with traditional CoA assisted diagnosis methods, 3DCoA model not only improves the reliability of diagnosis results, but also directly faces image operation without complicated data preprocessing process. The experimental results show that compared with the existing methods, the performance of 3DCoA model in diagnosis accuracy, precision and recall has been significantly improved.
  • Supplementary Material | Related Articles

  • Node importance evaluation in multiplex heterogeneous networks based on graph embedding
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 1-76.
  • Abstract ( 125 )     　　
  • Node importance evaluation is an effective means to analyze the structure and dynamic characteristics of complex networks. The existing node importance evaluation methods mainly focus on homogeneous networks. However, the most of networks in the real world is multiplex heterogeneous networks (MHEN), which evaluation of node importance has the challenges of multiple types of nodes and edges. A method of node importance evaluation is proposed for MHEN based on graph embedding. For the same type and different types of edges, the characteristics of the nodes are aggregated after random walk sampling neighbor nodes, so as to get the embedding vectors of nodes. The node importance evaluation index for MHEN is constructed by the embedding vectors of nodes and characteristics of local structure. The experimental results on four real network datasets show that compared with MBC, BPR and MEC methods, the proposed method performs better in the accuracy of node importance evaluation and overall ranking results.
  • Supplementary Material | Related Articles

  Special Topics on Intelligent Medical

  • Design of Neural Network Model for Auxiliary Diagnosis of Coarctation of Aorta
  • WU Xingkun, LUO Tao, LIU Aijun, YANG Ming, ZHANG Wenjing
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 1-6. DOI:10.13190/j.jbupt.2021-185
  • Abstract ( 354 )     HTML( 189 )     　　
  • A model of coarctation of aorta based on three-dimensional aided diagnosis model of aortic coarctation based on three-dimensional convolution neural network is proposed, which combines the three-dimensional spatial features of cardiac computed tomography images. Compared with the traditional auxiliary diagnosis method of aortic coarctation, the proposed method not only improves the reliability of diagnosis results, but also directly processes images operation without complicated data preprocessing process. The performance in terms of diagnosis accuracy, precision and recall has been significantly improved.
  • References | Supplementary Material | Related Articles

  • A Sleep Staging Method Combining Grouping Convolution with Semi-Supervised Learning
  • XIE Pan, PENG Caijing, HE Zhihui, ZHANG Yuan
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 7-12. DOI:10.13190/j.jbupt.2021-246
  • Abstract ( 358 )     HTML( 145 )     　　
  • Sleep is an important physiological activity of human body, and the quality of sleep affects physical and mental health. Most of the existing sleep staging is based on supervised learning, which is highly dependent on a large number of high-quality label data, and the extracted features are relatively rough. To save this issue, a sleep staging method combining grouped convolutional neural network and semi-supervised learning is proposed. First, the grouping residual convolution network is used as the backbone network to ensure the diversity of learning features and take the information from multiple subspaces into consideration, which extracts multi-angle features. Then, to reduce the workload of annotation technicians, semi-supervised learning method is adopted to extract features from a large number of unlabeled data and compete with those extracted from labeled data, which can obtain more fine-grained features. The experimental results show that the accuracy of sleep staging on sleep-EDFx can reach 0.837±0.001, and the Kappa coefficient reaches 0.774±0.002, which performs better than the baseline algorithm. The method presented has a good application prospect in the combination of medicine and industry.
  • References | Supplementary Material | Related Articles

  • Traditional Chinese Medicine Symptom Normalization Approach Based on Pre-Trained Language Models
  • XIE Yonghong, TAO Hu, JIA Qi, YANG Shibing, HAN Xinliang
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 13-18,57. DOI:10.13190/j.jbupt.2021-191
  • Abstract ( 439 )     HTML( 400 )     　　
  • To solve the issue in traditional Chinese medicine that one symptom has different literal descriptions and one symptom corresponds to multiple normalized descriptions, a two-stage framework based on pre-trained language models is proposed. In the first step, according to the definition and classification of symptoms, a multi-label text classification model is adopted to semantically divide the symptom descriptions to obtain candidate normalization symptom words. In the second step, we score and sort the candidate normalization symptom words with an entity matching model, and some strategies are designed to perform a second recall of the results to improve performance. After that, the candidate word with the highest score in each semantic label is regarded as the normalization result. Experiments results show that the proposed method performs better than traditional methods on solving the symptom normalization problem. Furthermore, the research compares and analyzes the results using different pre-trained language models on the symptom normalization task to verify the effectiveness of the proposed method.
  • References | Supplementary Material | Related Articles

  • Traditional Chinese Medicine Symptom Normalization Approach Based on Pre-trained Language Models
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 14-20.
  • Abstract ( 214 )     　　
  • Symptom normalization plays a vital role in mining Traditional Chinese medicine (TCM) knowledge and the promotion of the modernization of TCM. It is difficult because the challenges of symptom descriptions such as one symptom having different literal descriptions, one-to-many symptom descriptions. To deal with this problem, a two-stage framework based on pre-trained language models is proposed. First, a multi-label text classification model is adopted to semantically divide the symptom descriptions to obtain candidate normalization symptom words, according to the definition and classification of symptoms. Then score and sort the candidate words with a symptom word matching model, after which take the candidate word with the highest score in each semantic label as the normalization result of the symptom description. Finally, some strategies are designed to perform a second recall of the results to improve performance. The research analyzes the results obtained with different pre-trained models with a constructed symptom normalization dataset. The experiments show that the method and strategies can effectively deal with symptom normalization, among which the ERNIE-based model shows the best performance with F1 value 0.894.
  • Supplementary Material | Related Articles

  Special Topics on Intelligent Medical

  • Construction of Multi-Modal Knowledge Graph for Epilepsy Related Papers
  • LI Xingyuan, WANG Peng, SHEN Mu, LI Lei, ZHANG Lin
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 19-24. DOI:10.13190/j.jbupt.2021-187
  • Abstract ( 581 )     HTML( 338 )     　　
  • The performance of the existing named entity recognition and relation extraction models would sharply decline due to the lack of a large amount of annotated data for epilepsy-related papers. To solve this issue, a zero-resource named entity recognition and relation extraction model in the epilepsy domain is proposed based on medical data and a pre-training model from similar domains. The performance of the existing unsupervised and semi-supervised models on the epilepsy paper data set isevaluated, and then a domain adversarial network and a relation discriminator are introduced based on the characteristics of the data set to effectively improve the construction effect of the epilepsy domain knowledge graph. Electroencephalography (EEG) features of epilepsy patients are embedded into the knowledge graph in a visual modality. While improving the interpretability of EEG analysis, it builds a more intuitive multi-modal knowledge graph.
  • References | Supplementary Material | Related Articles

  • Construction of multi-modal knowledge graph in epilepsy
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 21-27.
  • Abstract ( 161 )     　　
  • Named entity recognition and relation extraction are two key steps to construct a knowledge graph. To solve the problem that the performance of the existing named entity recognition and relation extraction models would sharply decline due to the lack of a large amount of annotated data in the epilepsy domain, the named entity recognition and relation extraction models were improved according to the data characteristics of the papers in the epilepsy domain. A zero-resource named entity recognition and relation extraction model in the epilepsy domain is proposed based on medical data and a pre-training model from similar domains. The performance of the existing unsupervised and semi-supervised models on the epilepsy paper data set was evaluated, and then a domain adversarial network and a relation discriminator were introduced based on the characteristics of the data set to effectively improve the construction effect of the epilepsy domain knowledge graph. EEG features of epilepsy patients were embedded into the knowledge graph in a visual modality. While improving the interpretability of EEG analysis, it builds a more intuitive multi-modal knowledge graph.
  • Supplementary Material | Related Articles

  Special Topics on Intelligent Medical

  • Lung Nodule Detection System Based on Data Augmentation and Attention Mechanism
  • LI Yang, GAO Shiqi
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 25-30. DOI:10.13190/j.jbupt.2021-189
  • Abstract ( 756 )     HTML( 399 )     　　
  • To solve the problem of limited model learning ability caused by insufficient labeled medical image data and easy loss of tiny nodule features caused by sub-sampling in the process of deep detection, a lung computer aided-detection system based on a generative adversarial network based on computed tomography data augmentation and improved you only look once-V4 (YOLO-V4) detection framework is designed. First, the regularization method DropBlock is introduced into the nodule generation framework computed tomography-generative adversarial networks to augment the data of annotated medical images, which can improve the generation quality of pulmonary nodules. Second, the coordinate attention model is introduced in YOLO-V4, which was constructed to capture the position perception, direction perception and cross-channel information of pulmonary nodules, which can further help the model to detect the region of interest of pulmonary nodules more accurately. The experimental results show that the performance indexes of data augmentation and nodule detection in the lung nodule analysis 16 data set of the proposed lung computer aided detection system are superior to the comparison algorithm, which can effectively expand the data set and improve the performance of nodule detection.
  • References | Supplementary Material | Related Articles

  • A lung CAD system based on Data Augmentation and CA-YOLO-V4 detection framework
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 28-34.
  • Abstract ( 148 )     　　
  • For solving the problem of limited model learning ability caused by too little labeled medical image data and easy loss of tiny nodule features caused by sub-sampling in the process of deep detection, a lung CAD system based on CT-GAN data augmentation and improved YOLO-V4 detection framework was designed. In the first part, the regularization method DropBlock was introduced into the nodule generation algorithm CT-GAN to augment the data of annotated medical images, so as to improve the generation quality of pulmonary nodules.In the second part, attention model was introduced in YOLO-V4, and CA-YOLO-V4 framework was constructed to capture the position perception, direction perception and cross-channel information of pulmonary nodules, helped the model to detect the region of interest of pulmonary nodules more accurately. The experimental results show that the performance indexes of data augmentation and nodule detection in LUNA16 data set of the proposed lung CAD system are superior to the comparison algorithm, which can effectively expand the data set and improve the performance of nodule detection.
  • Supplementary Material | Related Articles

  Special Topics on Intelligent Medical

  • Classification of Mycosis Fungoides Cells Based on Multi Branch Squeeze and Excitation Network
  • WANG Junjie, XU Congcong, ZHAO Zengrui, XU Jun, JIANG Yiqun
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 31-36. DOI:10.13190/j.jbupt.2021-183
  • Abstract ( 309 )     HTML( 192 )     　　
  • To study the different cell components of mycosis fungoides, a multi branch squeeze and excitation network model is constructed based on 77 whole slide images of early and middle stage mycosis fungoides, and the classification of lymphocytes and epithelial cells of mycosis fungoides is realized. The network is divided into two stages:encoding and decoding. The encoding stage corresponds to one branch, and the decoding stage has three branches, corresponding to one main task and two auxiliary tasks. The main task branch outputs the results of cell classification, the auxiliary branch I outputs the cells and background, and the auxiliary branch II outputs the horizontal and vertical boundary map. In the training stage, 576 image blocks were selected from the slices and marked by professional pathologists, including 464 for training and 112 for verification. Finally, they are tested on the whole slide images. The cell segmentation accuracy and F1 score of the model are 0.943 and 0.728, respectively. The average accuracy of classification is 0.943. The experimental results show that the proposed model can recognize and classify lymphocytes and epithelial cells in mycosis fungoides, which lays an important foundation for computer-aided diagnosis of cutaneous mycosis fungoides.
  • References | Supplementary Material | Related Articles

  • Classification of mycosis fungoides cells based on multi branch Squeeze and Excitation network
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 35-41.
  • Abstract ( 157 )     　　
  • In order to study the different cell components of mycosis fungoides, a multi branch squeeze and excitation network model was constructed based on 77 whole slide images of early and middle stage mycosis fungoides, and the classification of lymphocytes and epithelial cells of mycosis fungoides was realized. The network is divided into two stages: encoding and decoding. The encoding stage corresponds to one branch, and the decoding stage has three branches, corresponding to one main task and two auxiliary tasks. The main task branch outputs the results of cell classification, the auxiliary branch I outputs the cells and background, and the auxiliary branch II outputs the horizontal and vertical boundary map. In the training stage, 576 image blocks were selected from the slices and marked by professional pathologists, including 464 for training and 112 for verification. Finally, they were tested on the whole slide images. The cell segmentation accuracy and F1 score of the model is 0.943 and 0.728, respectively. The average accuracy of classification is 0.943. The experimental results show that the proposed model can recognize and classify lymphocytes and epithelial cells in mycosis fungoides, which lays an important foundation for computer-aided diagnosis of cutaneous mycosis fungoides.
  • Supplementary Material | Related Articles

  Special Topics on Intelligent Medical

  • Medical Question-Answer Matching Base on Adversarial Training
  • FU Jieqiong, SUN Yawei, LIU Jianyi, LI Jinbin
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 37-43. DOI:10.13190/j.jbupt.2021-242
  • Abstract ( 447 )     HTML( 223 )     　　
  • Compared with the question-answer matching task in the English open domain, the task in the Chinese professional medical field is more challenging.In view of the complexity of Chinese semantics and the diversity of medical data,most researchers focus on designing complex neural networks to explore deeper text semantics, and this kind of idea is relatively simple.At the same time, the neural network model is easy to make misjudgments due to small disturbances, and the poor generalization ability of the model.To solve these issues,a question-answer matching model is proposed based on adversarial training. A bidirectional pre-training encoder is used in this model to capture the semantic information of question-answer sentences and to obtain the corresponding vector representation. Then, adversarial samples are generated by adding a disturbance factor to the word embedding representation.Finally, the initial samples and adversarial samples are jointly input into the model with linear layers for classification prediction.Comparative experiments demonstrate that adversarial training can effectively improve the performance of the question-answer matching model on the cMedQA V2.0 dataset.
  • References | Supplementary Material | Related Articles

  • Smoothing Attack Algorithm Based on Electrocardiogram Classification
  • LIU Jintong, YANG Guoxing, LIU Xiaohong, WANG Guangyu
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 44-50. DOI:10.13190/j.jbupt.2022-031
  • Abstract ( 395 )     HTML( 160 )     　　
  • In the field of electrocardiogram classification, the adversarial samples generated by the traditional projected gradient algorithm with low generation efficiency have square waves that cannot be explained physiologically, and thus, a patch-based smooth attack perturbations (PatchSAP) algorithm is proposed. By conducting adversarial attacks against three common electrocardiogram classification models, convolutional neural network, long-short-term memory network, and attention-based long-short-term memory network, we compare the "vulnerability" of the electrocardiogram classification models, and analyze the hyperparameter to obtain the difference between validity and authenticity of adversarial examples. The experimental results show that the PatchSAP algorithm has obvious advantages in attack efficiency, and the generated adversarial samples maintain the sample authenticity well. Hyperparameters such as convolution kernel and constraint range have a great impact on the effectiveness and authenticity of adversarial examples.
  • References | Supplementary Material | Related Articles

  • Automated Hippocampal Sclerosis Detection Method with Radiomics Analysis
  • OUYANG Mowei, KANG Guixia, WANG Kailiang, WANG Yaming
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 51-57. DOI:10.13190/j.jbupt.2022-069
  • Abstract ( 363 )     HTML( 118 )     　　
  • Based on radiomics analysis, an automatic diagnosis method of hippocampal sclerosis is proposed to find radiomics features to characterize hippocampal sclerosis lesions and improve the detection accuracy of hippocampal sclerosis. Combined with the wavelet transform and Laplacian of Gaussian filtering algorithm, the radiomics features in the region of hippocampus were extracted from several filtered images to obtain highly sensitive features and identify hippocampal sclerosis. First, pre-processing and filtering processes are applied to T1 image of tested samples, and radiomic features are extracted from the original image and the filtered image. Then, the student-T test and feature correlation analysis are used to reduce the dimension of the radiomic features. Finaly, hippocampal sclerosis detection models are built based on these features. The experimental results show that the hippocampal sclerosis detection model based on radiomics features can effectively assist the identification of hippocampal sclerosis lesions, and the radiomics features extracted from wavelet transform image have the best detection performance onthe automatic diagnosis model with a detection accuracy of 97.7% for the actual dataset.
  • References | Supplementary Material | Related Articles

  • Automatic Recognition of Mucus Impaction in CT Images of Asthmatic Patients Using Deep Learning
  • HUANG Liuting, LIU Kexin, NIU Kai, CHANG Chun, HE Zhiqiang
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 58-63. DOI:10.13190/j.jbupt.2022-030
  • Abstract ( 507 )     HTML( 330 )     　　
  • In view of the low efficiency of manual identification of mucus impaction in chest computed tomography (CT) and the poor recognition effect, a deep neural network based automatic recognition algorithm for mucus impaction is proposed. In order to deal with the irregular characteristics of mucus impaction, deformable convolution is added to the backbone to extract features, and deformable region of interest pooling is used in the detection network to normalize the feature scale. Besides, feature pyramid network with weight coefficient is used for multi-scale fusion according to the characteristics of small and medium objects. The results show that compared with the traditional faster region convolutional neural network, mean average precision of the proposed algorithm is improved by 4%, which can provide auxiliary reference for the diagnosis of asthma severity.
  • References | Supplementary Material | Related Articles

  Special Topics on Wireless Sensor Networks

  • Link Prediction in Opportunistic Networks Based on Network Representation Learning
  • LIU Linlan, SONG Xiuyang, CHEN Yubin
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 64-69,103. DOI:10.13190/j.jbupt.2021-211
  • Abstract ( 390 )     HTML( 190 )     　　
  • According to the characteristics of topology frequent changes and multi-dimensional attributes in opportunistic networks, a link prediction method based on network representation learning is proposed. The opportunistic network is transformed into snapshots by setting time slot. The link state of each snapshot is represented by multi-dimensional link attributes. Then, the network representation learning method is adopted to aggregate the multi-dimensional link attributes of neighbor nodes, which are mapped into a low-dimensional embedding matrix. The recurrent neural network improved based on the attention mechanism is employed to learn the laws of the evolution of network topology, and to extract the timing features between embedding matrices. Through the output layers, the mapping relationship between time serial characteristics and link-state is established to implement the link prediction for network at the next moment. The experimental results on mainstream datasets, such as Infocom-05 and Hyccups show that the proposed method achieves higher prediction accuracy compared with the existing link prediction methods.
  • References | Supplementary Material | Related Articles

  • User Fine-Grained Reliability and Truth Estimate Model on Mobile Crowdsensing
  • LIU Likun, QIU Tie, XU Tianyi, CHEN Ning, WAN Zhiguo
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 70-76. DOI:10.13190/j.jbupt.2021-217
  • Abstract ( 400 )     HTML( 108 )     　　
  • To improve perceived data quality in mobile crowdsensing, a method for estimating the truth value of crowdsensing tasks based on user fine-grained reliability is proposed, whichselect high-quality data through user modeling. First, the real-time reliability of users is evaluated according to the instantaneous factors that affect their task execution. Then, information entropy is introduced to measure the user's reputation distribution, including the user's overall reputation distribution and the reputation distribution under different tasks. Next, based on user reliability, an effective truth estimation method is designed to predict task truth. Experimental results show that the proposed model can effectively evaluate the reliability of users in multi-type tasks and improve the accuracy of task truth estimation.
  • References | Supplementary Material | Related Articles

  • Edge-Cloud Collaborative Worker Recruitment Algorithm in Mobile Crowd Sensing System
  • XI Heran, ZHU Jinghua, LI Jinbao
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 77-83. DOI:10.13190/j.jbupt.2021-205
  • Abstract ( 418 )     HTML( 282 )     　　
  • Since the recruitment algorithms based on cloud platform cannot meet the needs of large scale network real-time tasks, an edge-cloud collaboration recruitmentalgorithm is proposed whose aim is to reduce the data transmission delay and the energy consumption of intelligent devices. The cloud service layer performs task reception, division, release and result collection; the edge layer performs obtaining the real-time information of workers and constructing the recruitment model of workers,while the perceptual layer performs task propagation and data collection. The experimental results show that the proposed algorithm can not only meet the cost and time constraints, but also achieve good performance in space coverage and time by taking consideration of sensor type, worker quotation and the maximum number.
  • References | Supplementary Material | Related Articles

  • SDN-Based Integrated Convergent Network Routing Scheduling Mechanism
  • LI Jingbo, MA Li, MA Dongchao, FU Yingxun, LI Yang
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 84-90. DOI:10.13190/j.jbupt.2021-199
  • Abstract ( 435 )     HTML( 279 )     　　
  • To solve the problems of communication fluctuation, traffic load balancing and poor robustness of dynamic access equipment in integrated converged network. By combining the technology of software defined network (SDN), the accurate subjective and objective cost models are optimized. The index threshold G1 method is proposed as the subjective weighting method, and the standard deviation method is proposed as the objective weighting method. The multiplication integration is used for weighted combination to give the final link cost. A multi-path selection algorithm of integrated fusion network is proposed, and the optimized cost combination is applied to the multi-path Dijkstra variant algorithm to obtain the transmission link and combination cost. An integrated federated network routing topology and strategy are proposed, in which, different networks can choose multiple paths that meet their own characteristics, and can be forwarded in proportion. The results show that the proposed scheme optimizes differentiated path selection and traffic scheduling, increases link bandwidth utilization, and reduces network transmission delay.
  • References | Supplementary Material | Related Articles

  • Edge-Cloud Collaborative Worker Recruitment Algorithm in Mobile Crowd Sensing System
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 91-97.
  • Abstract ( 221 )     　　
  • This paper studies the problem of worker recruitment algorithm for mobile crowd sensing system (MCS). Recruitment algorithms based on cloud platform cannot meet the needs of large scale network real-time tasks, a three-tier worker recruitment algorithm(ECRecruitment) based on edge cloud collaboration is proposed, which aims to reduce the data transmission delay and the energy consumption of intelligent devices. The cloud service layer is responsible for task reception, division, release and result collection; The edge layer is responsible for obtaining the real-time information of workers and constructing the recruitment model of workers; The perceptual layer is responsible for task propagation and data collection. ECRecruitment considers multiple influencing factors, such as sensor type, worker quotation, maximum number of assigned tasks, etc. The experimental results show that ECRecruitment can not only meet the cost and time constraints, but also achieve good performance in space coverage and operation time.
  • Supplementary Material | Related Articles

  Special Topics on Wireless Sensor Networks

  • Multi-Objective Fusion Potential Game Wireless Ad Hoc Network Topology Control Algorithm
  • SU Yang, WEI Liansuo, GUO Yuan
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 91-97. DOI:10.13190/j.jbupt.2021-207
  • Abstract ( 411 )     HTML( 228 )     　　
  • To solve problems such as unbalanced load of individual "bottleneck nodes", many redundant links and short life cycle in existing topology control algorithm of wireless Ad hoc network based on game theory, a multi-objective fusion network topology control algorithm is proposed. First, by analyzing the influence of network connectivity, node transmission power, residual energy, node degree, link quality, and link length on node load, an improved and optimized comprehensive utility function is designed. Then, a multi-objective network topology control potential game model is established, and it is proved that the model is an ordinal potential game and has Nash equilibrium solution. On the basis of maintaining the k-connectivity of the network, the minimum path set algorithm is used to optimize the network topological links and eliminate redundant links after the gaming. Simulation experiments and comparative analysis show that the proposed algorithm can achieve network load balance and eliminate redundant links on the premise of ensuring network connectivity, and its life cycle is increased by 25.4%, 92.6%, and 36.8% compared with distributed topology control algorithm, energy balance topology control game algorithm and energy-efficient and fault-tolerant topology control game algorithm.
  • References | Supplementary Material | Related Articles

  • SDN-based integrated convergent network routing scheduling mechanism
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 98-104.
  • Abstract ( 251 )     　　
  • In recent years, the Internet of Everything has developed rapidly, and a large number of network protocols and types that meet different needs have emerged. In the scenario of integrated network requirements, the network integrating different types and protocols has problems such as unstable communication, unbalanced data traffic load, and poor robustness. Software Defined Networking (SDN), as a new type of network paradigm, provides a convenient way to solve these problems. Combined with SDN analysis and research, the dynamic mechanism is used to solve the routing and scheduling problem of the integrated convergent network. First, the link information is divided into 8 combinations of hop count, delay, bandwidth, and packet loss rate. Different networks select different combination costs according to their characteristics, and monitor and calculate network information through real-time traffic modules. Secondly, the accurate cost model based on subjective and objective is optimized. Based on the subjective weighting method G1 method, the index threshold type G1 method is proposed and then the multiplicative integration is used for weighted combination to give the final link cost. At the same time, an integrated multi-path selection algorithm for converged networks is designed, and the optimized cost combination is applied to the multi-path Dijkstra variant algorithm to obtain the transmission link and the combined cost. According to the routing strategy, the multi-path that meets its own characteristics is selected for different networks and proportionally forward. The routing and scheduling mechanism in SDN is simulated through the Ryu controller and Mininet platform. The results show that, through the dynamic and multi-dimensional full-time air conditioning mechanism, the multi-path characteristics of the integrated converged network are fully explored, and the differentiated path selection and flow scheduling of the integrated converged network in a dynamic environment are realized, and the delay and bandwidth utilization Obvious results have been achieved on multiple indicators such as transmission rate, jitter, and packet loss rate.
  • Supplementary Material | Related Articles

  Special Topics on Wireless Sensor Networks

  • Combined Multi-Resource Task Offloading Algorithm for P2P in Edge Computing
  • LU Weifeng, LI Xueqing, XU Jia, CHEN Siguang
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 98-103. DOI:10.13190/j.jbupt.2021-200
  • Abstract ( 317 )     HTML( 137 )     　　
  • In order to solve the problem of task offloading that requires the cooperation of multiple resources in the peer-to-peer system, a multi-resource combination transaction offloading algorithm is proposed, and an incentive mechanism is designed to encourage devices to join the task offloading system, while ensuring that the resources in the system can be efficiently utilized. Through rigorous theoretical analysis, it is proved that the designed multi-resource combination transaction mechanism satisfies the feasibility of calculation efficiency and individual rationality, and extensive experimental simulations are carried out, although the number of resource transactions of the proposed multi-resource combination transaction offloading algorithm does not reach the number of resource transactions under the comparison algorithm, but the time complexity of the algorithm is far lower than the comparison algorithm.
  • References | Supplementary Material | Related Articles

  • Node Importance Evaluation in Multiplex Heterogeneous Network Based on Graph Embedding
  • SHU Jian, YAO Xiaolong, LI Ruirui
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 104-109. DOI:10.13190/j.jbupt.2021-214
  • Abstract ( 431 )     HTML( 246 )     　　
  • To improve the accuracy of node importance evaluation in multiplex heterogeneous network (MHEN), a method of node importance evaluation is proposed for MHEN based on graph embedding. For the same type and different types of edges, the features of the nodes are aggregated after random walk sampling neighbor nodes, and the features are mapped to the embedding space by multi-layer perceptron to obtain the embedding vectors. Then, the node importance evaluation index for MHEN is constructed by the embedding vectors of nodes and features of local structure. The experimental results on mainstream datasets, such as CElegans and CS-Aarhus show that compared with multiplex betweenness centrality, biased PageRank and multiplex evidential centrality, the proposed method performs better in term of the accuracy.
  • References | Supplementary Material | Related Articles

  • Multi-Objective Fusion Ordinal Potential Game Wireless Ad Hoc Network Topology Control Algorithm
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 105-111.
  • Abstract ( 187 )     　　
  • Aiming at the problems of unbalanced load of individual "bottleneck nodes", multiple redundant links, and short life cycle in the existing wireless ad hoc network topology control algorithms based on game theory, this paper proposes a multi-objective fusion ordinal potential game wireless auto Group network topology control algorithm. The algorithm first designs an improved and optimized comprehensive utility function by analyzing the influence of network connectivity, node transmit power, remaining energy, node degree, link quality, and link length on node load； based on this, a multi-objective is established Converged network topology control potential game model, and proved that the model is an ordinal potential game and has a Nash equilibrium solution； on the basis of maintaining the network k-connection, the minimum path set algorithm is used to optimize the network topology links after the game, and eliminate Redundant link. Simulation experiments and comparative analysis show that the algorithm achieves network load balancing and eliminates redundant links under the premise of ensuring network connectivity. Compared with DEBA algorithm, PGTC algorithm and EBTG algorithm, the life cycle is increased by 23. 3%, 68. 8%, 98. 5%.
  • Supplementary Material | Related Articles

  Reports

  • Automatic Generation Method of Turtle Back Pattern Based on Mathematical Rules
  • ZHAO Haiying, XIE Guangpeng, GAO Zihui
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 110-115. DOI:10.13190/j.jbupt.2021-100
  • Abstract ( 420 )     HTML( 216 )     　　
  • Turtle back pattern is a kind of Chinese folk decorative pattern, generally hexagonal continuous geometric pattern. By analyzing the rules of turtle back pattern composition, a new algorithm of traditional folk pattern generation is proposed. Firstly, the most basic elements of turtle ridges (called primitives) are analyzed, and the basic elements of primitives are obtained according to the symmetry of different primitives. Then by rotating the basic elements to obtain primitives, and tiling primitives to obtain the entire turtle back pattern. By defining symbols and parameterizing the generating process of primions, a general generating formula of tortoise back pattern is obtained. The simulation results show that the tortoisesback patterns with different details can be generated by adjusting the parameters of the generating formula.
  • References | Supplementary Material | Related Articles

  • Siamese Network Target Trackingthat Combines Dual Attention and Feature Fusion
  • LI Xue, LI Xiaoyan, WANG Peng, SUN Mengyu, Lü Zhigang
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 116-122. DOI:10.13190/j.jbupt.2021-169
  • Abstract ( 786 )     HTML( 479 )     　　
  • In order to solve the problem that the evolution of siamese visual tracking with very deep networks (SiamRPN++) algorithm when the target is occluded and deformed. A siamese network target tracking algorithm combining dual attention and feature fusion is proposed. Firstly, the channel and spatial attention module are used to enhance the target information, suppress the interference information in the image and improve the accuracy of the model; then, multi-layer feature fusion is carried out for the shallow and deep feature information output from the attention layer to obtain better expressive target features and improve the tracking accuracy; finally, the online template update mechanism is introduced to reduce tracking drift and improve the tracking robustness. The OTB100 dataset is used for experimental and the results show that the success rate of the improved algorithm is increased by 1.3% compared with SiamRPN++, indicating that the tracking accuracy of the algorithm is higher; under the four test sequences with occlusion and deformation attributes, the average overlap rate of the improved algorithm is increased by 3%, and the average center position error is reduced by 0.37 pixels, which is better robustness against occlusion and appearance deformation.
  • References | Supplementary Material | Related Articles

  • Automatic generation method of turtle back pattern based on mathematical rules
  • Hai YingZhao Zi huiGAO
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 119-125.
  • Abstract ( 138 )     　　
  • This paper presents an algorithm for generating turtle back pattern, a traditional national pattern. The algorithm is based on the analysis of the mathematical rules of turtle back pattern. It mainly analyzes the motif of turtle back pattern, and then obtains the primitive elements of the motif from the symmetry of the primitive. After the rotation and symmetry of the primitive elements, the motif is obtained, and the whole turtle back pattern is obtained by tiling the motif. Moreover, through the formulation of the primitive elements, we get the generalization generation formula of turtle back pattern. By introducing certain parameters, we can generate a large number of similar but different patterns. The experimental results also verify the feasibility of this method, which is of great significance for the digital preservation of national traditional patterns.
  • Supplementary Material | Related Articles

  Reports

  • Cell Image Segmentation Algorithm Based on Minimum Operation of Color Model
  • LEI Yu, ZHANG Lijuan, TANG Peng, HU Miao, LI Dongming
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 123-128. DOI:10.13190/j.jbupt.2021-219
  • Abstract ( 313 )     HTML( 186 )     　　
  • To detect the morphology, proliferation, and differentiation of human cells quickly and accurately, an efficient adaptive cell image segmentation method is proposed. First, the image of saturation with obvious characteristics is extracted from the Hue,Saturation Value (HSV) channel of the image, and then, morphological reconstruction, H-minima technology, and image enhancement technology are used to perform gradient correction on the image of saturation. After the gradient of the image is corrected, the watershed algorithm is used to perform segmentation. Next, the segmentation result is merged to obtain the background, cell, and nucleus according to the gray level consistency of the original image. Finally, post-morphological processing is used to remove the false noise in the merged result and flatten the edge of the region. The results of the segmentation test on cell images show that the accuracies of cell and cell nucleus segmentation are 0.978 8 and 0.967 7, respectively. The Dice coefficients are 0.938 8 and 0.937, which realizes the accurate segmentation of medical micro cell images.
  • References | Supplementary Material | Related Articles

  • Cell Image Segmentation Algorithm Based on Minimum Operation of Color Model
  • Journal of Beijing University of Posts and Telecommunications. 2022, 45(4): 133-139.
  • Abstract ( 149 )     　　
  • To be able to quickly and accurately detect the morphology, proliferation, and differentiation of human cells, and evaluate the health of patients, an efficient adaptive cell image segmentation method based on the minimum operation of the color model is proposed. First, extract the S-component image with obvious characteristics in the HSV channel of the image, and use morphological reconstruction, H-minima technology, and image enhancement technology to perform gradient correction on the S-component image; after the gradient of the image is corrected, use the watershed algorithm to perform Segmentation: After the segmented image is obtained, the segmentation result is merged according to the gray level consistency of the original image to obtain the background, cell, and nucleus. Finally, post-morphological processing is used to remove the false noise in the merged result and flatten the edge of the region. The results of the segmentation test on cell images show that the accuracy of cell and cell nucleus segmentation is 0.9788 and 0.9677, respectively. The Dice coefficients are 0.9388 and 0.9370, which realizes the accurate segmentation of medical micro cell images.
  • Supplementary Material | Related Articles