Email alert
2021 Vol. 43, No. 4
Display Method:
2021, 43(4): 881-891.
doi: 10.11999/JEIT200158
Abstract:
The security of cloud data is one of the most important factors to obstruct the development of cloud computing. Therefore, on the basis of proposed three-tiers cloud architecture that consists of physical resources layer, virtual component layer, and cloud service layer, this paper makes a detail survey on existing works that focus on the security of cloud data, which involves in cloud identify authentication, cloud access control, cloud data secure computing, virtualization, cloud data security storage, cloud data secure deletion, information flow control, cloud data secure auditing, cloud data privacy preserving, and cloud business continuity, respectively. Finally, research trends in the field of cloud data security are presented.
The security of cloud data is one of the most important factors to obstruct the development of cloud computing. Therefore, on the basis of proposed three-tiers cloud architecture that consists of physical resources layer, virtual component layer, and cloud service layer, this paper makes a detail survey on existing works that focus on the security of cloud data, which involves in cloud identify authentication, cloud access control, cloud data secure computing, virtualization, cloud data security storage, cloud data secure deletion, information flow control, cloud data secure auditing, cloud data privacy preserving, and cloud business continuity, respectively. Finally, research trends in the field of cloud data security are presented.
2021, 43(4): 892-899.
doi: 10.11999/JEIT191046
Abstract:
Users’ offloading tasks and offloading frequencies in Mobile Edge Computing(MEC) may cause users to be locked out. A privacy-preserving computation offloading method based on k-anonymity is proposed in this paper. Firstly, based on the differences between offloading tasks and their frequencies, privacy constraint is proposed to establish a privacy-preserving computation offloading model based on offloading frequency; Then, a Privacy-preserving Computation Offloading algorithm based on Simulated Annealing (PCOSA) is utilized to obtain the optimal k-anonymous groups and the privacy constraint frequency of each task; Finally, the user’s original offloading frequencies are changed to meet the privacy constraint while minimizing terminal energy consumption. Simulation results validate that the PCOSA can find out k users with the closest offloading performance to form anonymous sets, which protects effectively the privacy of all users.
Users’ offloading tasks and offloading frequencies in Mobile Edge Computing(MEC) may cause users to be locked out. A privacy-preserving computation offloading method based on k-anonymity is proposed in this paper. Firstly, based on the differences between offloading tasks and their frequencies, privacy constraint is proposed to establish a privacy-preserving computation offloading model based on offloading frequency; Then, a Privacy-preserving Computation Offloading algorithm based on Simulated Annealing (PCOSA) is utilized to obtain the optimal k-anonymous groups and the privacy constraint frequency of each task; Finally, the user’s original offloading frequencies are changed to meet the privacy constraint while minimizing terminal energy consumption. Simulation results validate that the PCOSA can find out k users with the closest offloading performance to form anonymous sets, which protects effectively the privacy of all users.
2021, 43(4): 900-907.
doi: 10.11999/JEIT200083
Abstract:
The Attribute-Based Encryption (ABE) mechanism is one of the key technologies for implementing flexible access control of data in cloud storage. However, the existing ABE schemes have some problems, such as too much ciphertext storage overhead and user privacy leakage, and unsupported public auditing of cloud data. To solve these problems, a new searchable ABE scheme is proposed, and its security can be reduced to the difficulty of q-BDHE (q –decisional Bilinear Diffie-Hellman Exponent) problem and CDH (Computational Diffie-Hellman) problem. The proposed scheme achieves a constant ciphertext length on the basis of supporting keyword search. By introducing strategies to hide ideas, it prevents attackers from obtaining sensitive information and ensures the privacy of users. And the integrity of the data in cloud storage is verified through data public audit mechanism. Compared with the existing similar schemes, this scheme greatly reduces the data encryption overhead, keyword search overhead, ciphertext storage cost and decryption cost, which has a good application prospect to the cloud storage environment.
The Attribute-Based Encryption (ABE) mechanism is one of the key technologies for implementing flexible access control of data in cloud storage. However, the existing ABE schemes have some problems, such as too much ciphertext storage overhead and user privacy leakage, and unsupported public auditing of cloud data. To solve these problems, a new searchable ABE scheme is proposed, and its security can be reduced to the difficulty of q-BDHE (q –decisional Bilinear Diffie-Hellman Exponent) problem and CDH (Computational Diffie-Hellman) problem. The proposed scheme achieves a constant ciphertext length on the basis of supporting keyword search. By introducing strategies to hide ideas, it prevents attackers from obtaining sensitive information and ensures the privacy of users. And the integrity of the data in cloud storage is verified through data public audit mechanism. Compared with the existing similar schemes, this scheme greatly reduces the data encryption overhead, keyword search overhead, ciphertext storage cost and decryption cost, which has a good application prospect to the cloud storage environment.
2021, 43(4): 908-914.
doi: 10.11999/JEIT200234
Abstract:
ACE as an authenticated encryption algorithm is selected as one of the round 2 candidates of the lightweight crypto standardization process. Since its excellent design advantages, e.g. simple structure, high performance in software and hardware, and suitable for constrained environments, the security of ACE is received extensive attention. In this paper, the concept of word propagation trail is introduced, and an exact model is constructed to describe the trail. A new automatic method for evaluating the security of word-based cipher against the integral attack is also proposed by using this model. Moreover, based on the structure of ACE, the security of ACE permutation is evaluated by using this new automatic method. More specifically, a new 12-step integral distinguisher of ACE permutation is verified by using this method, which requires the data complexity of about 2256 chosen data, the time complexity of about 2256 12-step ACE permutation operations, and the memory complexity of about 8 Byte. Compared with the distinguishers given by ACE’s designer, this new result prominently increases 4 steps indeed.
ACE as an authenticated encryption algorithm is selected as one of the round 2 candidates of the lightweight crypto standardization process. Since its excellent design advantages, e.g. simple structure, high performance in software and hardware, and suitable for constrained environments, the security of ACE is received extensive attention. In this paper, the concept of word propagation trail is introduced, and an exact model is constructed to describe the trail. A new automatic method for evaluating the security of word-based cipher against the integral attack is also proposed by using this model. Moreover, based on the structure of ACE, the security of ACE permutation is evaluated by using this new automatic method. More specifically, a new 12-step integral distinguisher of ACE permutation is verified by using this method, which requires the data complexity of about 2256 chosen data, the time complexity of about 2256 12-step ACE permutation operations, and the memory complexity of about 8 Byte. Compared with the distinguishers given by ACE’s designer, this new result prominently increases 4 steps indeed.
Heterogeneous Information Network Representation Learning Framework Based on Graph Attention Network
2021, 43(4): 915-922.
doi: 10.11999/JEIT200034
Abstract:
Commonly used heterogeneous information networks include knowledge graphs and heterogeneous information networks with simple schemas. Their representation learning follows usually different methods. The similarities and differences between knowledge graphs and heterogeneous information networks with simple schemas are summarized, and a general heterogeneous information network representation learning framework is proposed. The proposed framework can be divided into three parts: the basic vector model, the graph attention network based propagation model, and the task model. The basic vector model is used to learn basic network vector; The propagation model learns the high-order neighbor features of the network by stacking attention layers. The replaceable task module is suitable for different application scenarios. Compared with the benchmark model, the proposed framework achieves relatively good results in the link prediction task of the knowledge graph and the node classification task of the heterogeneous information network.
Commonly used heterogeneous information networks include knowledge graphs and heterogeneous information networks with simple schemas. Their representation learning follows usually different methods. The similarities and differences between knowledge graphs and heterogeneous information networks with simple schemas are summarized, and a general heterogeneous information network representation learning framework is proposed. The proposed framework can be divided into three parts: the basic vector model, the graph attention network based propagation model, and the task model. The basic vector model is used to learn basic network vector; The propagation model learns the high-order neighbor features of the network by stacking attention layers. The replaceable task module is suitable for different application scenarios. Compared with the benchmark model, the proposed framework achieves relatively good results in the link prediction task of the knowledge graph and the node classification task of the heterogeneous information network.
2021, 43(4): 923-930.
doi: 10.11999/JEIT190992
Abstract:
Class imbalance always exists in the process of network traffic classification. Considering the problem, a new feature selection algorithm using Weighted Symmetric Uncertainty (WSU) and Approximate Markov Blanket (AMB) is proposed. Firstly, a feature metric is defined using category distribution information, which is biased to minority classes. This makes it easier pick out features which have strong correlation with minority classes. Then, considering the correlation between features and categories and between features and features, the weighted symmetry uncertainty and approximate Markov blanket are used to delete the unrelated features and redundant features. Finally, the feature dimension is further reduced to determine the optimal feature subset, by using feature evaluation functions based on correlation measures and sequence search algorithms. The experimental results demonstrate that the algorithm can effectively improve the classification performance of minority classes without sacrificing the accuracy of the overall classification.
Class imbalance always exists in the process of network traffic classification. Considering the problem, a new feature selection algorithm using Weighted Symmetric Uncertainty (WSU) and Approximate Markov Blanket (AMB) is proposed. Firstly, a feature metric is defined using category distribution information, which is biased to minority classes. This makes it easier pick out features which have strong correlation with minority classes. Then, considering the correlation between features and categories and between features and features, the weighted symmetry uncertainty and approximate Markov blanket are used to delete the unrelated features and redundant features. Finally, the feature dimension is further reduced to determine the optimal feature subset, by using feature evaluation functions based on correlation measures and sequence search algorithms. The experimental results demonstrate that the algorithm can effectively improve the classification performance of minority classes without sacrificing the accuracy of the overall classification.
2021, 43(4): 931-938.
doi: 10.11999/JEIT200880
Abstract:
Existing segmentation based methods have problems, such as the difficulty in distinguishing adjacent text areas and the low efficiency of model detection caused by the complex steps in the post-processing stage. In order to solve this problem, this article proposes a novel scene text detection model based on fully convolutional network, which can solve the problem that adjacent texts are difficult to distinguish in existing methods and improve the detection speed of the model. First, it constructs a feature extractor to extract multi-scale feature map from the input image. Secondly, the bidirectional feature fusion module is used to fuse the semantic information of the two parallel branches and promote the joint optimization of the two branches. It then effectively differentiates adjacent texts by predicting both a reduced text area map and a full text area map in parallel. The former can guarantee the distinction between different text instances, while the latter can effectively guide the network optimization. Finally, in order to improve the speed of text detection, it proposes a fast and effective post-processing algorithm to generate text boundary boxes. The experimental results show that: on relative datasets, the method proposed in this article achieves the best performance, and improves the F-measure index by 1.0% at most compared with the current best method, and can achieve near-real-time speed, which proves fully the effectiveness and high efficiency of the method.
Existing segmentation based methods have problems, such as the difficulty in distinguishing adjacent text areas and the low efficiency of model detection caused by the complex steps in the post-processing stage. In order to solve this problem, this article proposes a novel scene text detection model based on fully convolutional network, which can solve the problem that adjacent texts are difficult to distinguish in existing methods and improve the detection speed of the model. First, it constructs a feature extractor to extract multi-scale feature map from the input image. Secondly, the bidirectional feature fusion module is used to fuse the semantic information of the two parallel branches and promote the joint optimization of the two branches. It then effectively differentiates adjacent texts by predicting both a reduced text area map and a full text area map in parallel. The former can guarantee the distinction between different text instances, while the latter can effectively guide the network optimization. Finally, in order to improve the speed of text detection, it proposes a fast and effective post-processing algorithm to generate text boundary boxes. The experimental results show that: on relative datasets, the method proposed in this article achieves the best performance, and improves the F-measure index by 1.0% at most compared with the current best method, and can achieve near-real-time speed, which proves fully the effectiveness and high efficiency of the method.
2021, 43(4): 939-947.
doi: 10.11999/JEIT190928
Abstract:
Iris recognition faces two important issues. they are how to decompose finely and reconstruct the spherical image of the iris, and how to identify the characteristics of the iris. Conventional iris recognition uses usually the planar features of these iris images. However, the human eye is a sphere. The geometric position information of the iris surface is an important signal, but it is difficult to extract the geometric features of the iris sphere from the planar image. Considering the issue that the plane features are prone to distortion and lack fidelity of iris texture, an Orthogonal and Symmetric Spherical Haar Wavelet (OSSHW) basis is proposed to decompose and reconstruct the spherical iris signal to obtain stronger geometric features of iris surface. The comparison of the feature extraction ability to spherical signal by the spherical harmonics and the typical semi-orthogonal or nearly orthogonal spherical Haar wavelet is also presented. And then, an iris recognition method based on Convolutional Neural Networks (CNN) + OSSHW is proposed, which can effectively capture the local fine features of iris spherical surface, and has stronger ability in iris recognition than semi-orthogonal or nearly orthogonal spherical Haar wavelet bases.
Iris recognition faces two important issues. they are how to decompose finely and reconstruct the spherical image of the iris, and how to identify the characteristics of the iris. Conventional iris recognition uses usually the planar features of these iris images. However, the human eye is a sphere. The geometric position information of the iris surface is an important signal, but it is difficult to extract the geometric features of the iris sphere from the planar image. Considering the issue that the plane features are prone to distortion and lack fidelity of iris texture, an Orthogonal and Symmetric Spherical Haar Wavelet (OSSHW) basis is proposed to decompose and reconstruct the spherical iris signal to obtain stronger geometric features of iris surface. The comparison of the feature extraction ability to spherical signal by the spherical harmonics and the typical semi-orthogonal or nearly orthogonal spherical Haar wavelet is also presented. And then, an iris recognition method based on Convolutional Neural Networks (CNN) + OSSHW is proposed, which can effectively capture the local fine features of iris spherical surface, and has stronger ability in iris recognition than semi-orthogonal or nearly orthogonal spherical Haar wavelet bases.
2021, 43(4): 948-955.
doi: 10.11999/JEIT200043
Abstract:
The algorithm of Bisecting K-means obtains multiple clustering results by using a set of initial center pairs to segment a cluster, and then selects the best from them to mitigate the adverse effect of the local optimal convergence on the performance of the algorithm. However, the current methods of random sampling to generate initial center pairs for Bisecting K-means have some problems, such as low efficiency, poor stability, missing values and so on, which are not competent for big data clustering. In order to solve these problems, firstly the lower triangular matrix composed by the pairs of initial centers and the lower triangular matrix composed by serial numbers of the pairs of initial centers are created. Then, by establishing several mappings between the elements and their positions in the two matrices, a linear complexity algorithm is proposed to generate initial center pairs from the set of random integers. Both theoretical analysis and experimental results show that the time efficiency and efficiency stability of this method are significantly better than the current methods of random sampling, so it is particularly suitable for these scenarios of high-dimensional big data clustering.
The algorithm of Bisecting K-means obtains multiple clustering results by using a set of initial center pairs to segment a cluster, and then selects the best from them to mitigate the adverse effect of the local optimal convergence on the performance of the algorithm. However, the current methods of random sampling to generate initial center pairs for Bisecting K-means have some problems, such as low efficiency, poor stability, missing values and so on, which are not competent for big data clustering. In order to solve these problems, firstly the lower triangular matrix composed by the pairs of initial centers and the lower triangular matrix composed by serial numbers of the pairs of initial centers are created. Then, by establishing several mappings between the elements and their positions in the two matrices, a linear complexity algorithm is proposed to generate initial center pairs from the set of random integers. Both theoretical analysis and experimental results show that the time efficiency and efficiency stability of this method are significantly better than the current methods of random sampling, so it is particularly suitable for these scenarios of high-dimensional big data clustering.
2021, 43(4): 956-964.
doi: 10.11999/JEIT200061
Abstract:
The Stochastic Gradient Descent (SGD) algorithm randomly picks up a sample to estimate gradients, creating big variance which reduces the convergence speed and makes the training unstable. A Distributed SGD based on Average variance reduction, called DisSAGD is proposed. The method uses the average variance reduction based on historical gradients to update parameters in the machine learning model, requiring little gradient calculation and additional storage, but using the asynchronous communication protocol to share parameters across nodes. In order to solve the “update staleness” problem of global parameter distribution, a learning rate with an acceleration factor and an adaptive sampling strategy are included: on the one hand, when the parameter deviates from the optimal value, the acceleration factor is increased to speed up the convergence; on the other hand, when one work node is faster than the other ones, more samples are sampled for the next iteration, so that the node has more time to calculate the local gradient. Experiments show that the DisSAGD reduces significantly the waiting time of loop iterations, accelerates the convergence of the algorithm being faster than that of the controlled methods, and obtains almost linear acceleration in distributed cluster environments.
The Stochastic Gradient Descent (SGD) algorithm randomly picks up a sample to estimate gradients, creating big variance which reduces the convergence speed and makes the training unstable. A Distributed SGD based on Average variance reduction, called DisSAGD is proposed. The method uses the average variance reduction based on historical gradients to update parameters in the machine learning model, requiring little gradient calculation and additional storage, but using the asynchronous communication protocol to share parameters across nodes. In order to solve the “update staleness” problem of global parameter distribution, a learning rate with an acceleration factor and an adaptive sampling strategy are included: on the one hand, when the parameter deviates from the optimal value, the acceleration factor is increased to speed up the convergence; on the other hand, when one work node is faster than the other ones, more samples are sampled for the next iteration, so that the node has more time to calculate the local gradient. Experiments show that the DisSAGD reduces significantly the waiting time of loop iterations, accelerates the convergence of the algorithm being faster than that of the controlled methods, and obtains almost linear acceleration in distributed cluster environments.
2021, 43(4): 965-973.
doi: 10.11999/JEIT191053
Abstract:
This paper proposes a multi-focus image fusion algorithm based on super pixel-level Convolutional Neural Network (sp-CNN). In this method, multi-scale super pixel segmentation is firstly applied to the source image to obtain the super pixels. Secondly, the sp-CNN is proposed to acquire the initial decision maps. Thirdly, according to the similarities and differences of the multiple initial decision maps, the uncertain region is reclassified by spatial frequency to obtain the phase decision map. At last, the final decision map is achieved to fuse the source images by post-processing the phase decision graph with morphology. Experimental results show that the proposed method achieves the goal of reducing time complexity and attains better fusion effect compared with the state-of-the-art fusion methods which utilize overlapping blocks.
This paper proposes a multi-focus image fusion algorithm based on super pixel-level Convolutional Neural Network (sp-CNN). In this method, multi-scale super pixel segmentation is firstly applied to the source image to obtain the super pixels. Secondly, the sp-CNN is proposed to acquire the initial decision maps. Thirdly, according to the similarities and differences of the multiple initial decision maps, the uncertain region is reclassified by spatial frequency to obtain the phase decision map. At last, the final decision map is achieved to fuse the source images by post-processing the phase decision graph with morphology. Experimental results show that the proposed method achieves the goal of reducing time complexity and attains better fusion effect compared with the state-of-the-art fusion methods which utilize overlapping blocks.
2021, 43(4): 974-981.
doi: 10.11999/JEIT200031
Abstract:
A novel method for Digital Surface Model (DSM) reconstruction of single-view remote sensing image is proposed which only relies on light detection and ranging data. Based on deep learning technology, a semantic segmentation network with an encode-decode structure is designed. The network uses Multi-scale Residual Fusion Encode and Decode (MRFED) blocks to extract semantic information from the input image, and then predicts the height value pixel by pixel, as well as adopts a strategy of skip connections with feature maps to preserves the detailed features and structural information of the input image. The model is trained and tested on a public dataset of remote sensing images containing DSM data. Experiments show that, the Mean Absolute Error (MAE) between DSM reconstruction results and true values is 2.1e-02, the Root Mean Square Error (RMSE) is 3.8e-02, and the Structural SIMilarity (SSIM) is 92.89%, which are all better than the classic deep learning semantic segmentation networks. Experiments confirm that the method can effectively reconstruct the DSM of single-view remote sensing images with high accuracy, as well as the structure of feature distribution.
A novel method for Digital Surface Model (DSM) reconstruction of single-view remote sensing image is proposed which only relies on light detection and ranging data. Based on deep learning technology, a semantic segmentation network with an encode-decode structure is designed. The network uses Multi-scale Residual Fusion Encode and Decode (MRFED) blocks to extract semantic information from the input image, and then predicts the height value pixel by pixel, as well as adopts a strategy of skip connections with feature maps to preserves the detailed features and structural information of the input image. The model is trained and tested on a public dataset of remote sensing images containing DSM data. Experiments show that, the Mean Absolute Error (MAE) between DSM reconstruction results and true values is 2.1e-02, the Root Mean Square Error (RMSE) is 3.8e-02, and the Structural SIMilarity (SSIM) is 92.89%, which are all better than the classic deep learning semantic segmentation networks. Experiments confirm that the method can effectively reconstruct the DSM of single-view remote sensing images with high accuracy, as well as the structure of feature distribution.
2021, 43(4): 982-991.
doi: 10.11999/JEIT200419
Abstract:
To deal with the problem of missing information caused by zero-shot image classification during building a shared attribute layer, a compensation method is proposed to embed the attribute correlation. The proposed zero-shot classification utilizes Semantic AautoEncoder (SAE) to realize the feature-to-attribute mapping, and the invisible images are classified using maximum posterior probability estimation based on the class Gaussian distribution model. In order to make up for the lack of attribute relationships in SAE learning, the additive and multiplicative factors are introduced to embed the attribute correlation. The particle swarm algorithm is used to search for the optimal factor parameters to achieve the compensation of attribute correlation information. Experimental results show that when the same mapping method is adopted, the classification performance of zero-shot image classification based on attribute correlation on Pubfig and OSR data sets is significantly improved compared with other methods.
To deal with the problem of missing information caused by zero-shot image classification during building a shared attribute layer, a compensation method is proposed to embed the attribute correlation. The proposed zero-shot classification utilizes Semantic AautoEncoder (SAE) to realize the feature-to-attribute mapping, and the invisible images are classified using maximum posterior probability estimation based on the class Gaussian distribution model. In order to make up for the lack of attribute relationships in SAE learning, the additive and multiplicative factors are introduced to embed the attribute correlation. The particle swarm algorithm is used to search for the optimal factor parameters to achieve the compensation of attribute correlation information. Experimental results show that when the same mapping method is adopted, the classification performance of zero-shot image classification based on attribute correlation on Pubfig and OSR data sets is significantly improved compared with other methods.
2021, 43(4): 992-1002.
doi: 10.11999/JEIT200033
Abstract:
The glioma detection and focus segmentation in Magnetic Resonance Imaging (MRI) has important value for the therapeutic schedule selection and the surgical operations. In order to improve the detection efficiency and segmentation accuracy for glioma, this paper proposes a two-stage calculating method. First, a light convolutional neural network is designed to implement rapidly detection and localization for the glioma in MR images. Then, the peritumoral edema, non-enhancing tumor, enhancing tumor, and normal are classified and segmented from each other through an Ensemble Learning (EL) process. In order to improve the accuracy of segmentation, 416 radiomics features extracted from multi-modal MR images and 128 CNN features extracted by a convolutional neural network are mixed. The feature vector consisting of 298 features for classification learning are formed after a feature reduction process. In order to verify the performance of the proposed algorithm, experiments are carried out on the BraTS2017 dataset. The experimental results show that the proposed method can quickly detect and locate the tumor. The overall segmentation accuracy is improved distinctly with respect to 4 state-of-the-art approaches.
The glioma detection and focus segmentation in Magnetic Resonance Imaging (MRI) has important value for the therapeutic schedule selection and the surgical operations. In order to improve the detection efficiency and segmentation accuracy for glioma, this paper proposes a two-stage calculating method. First, a light convolutional neural network is designed to implement rapidly detection and localization for the glioma in MR images. Then, the peritumoral edema, non-enhancing tumor, enhancing tumor, and normal are classified and segmented from each other through an Ensemble Learning (EL) process. In order to improve the accuracy of segmentation, 416 radiomics features extracted from multi-modal MR images and 128 CNN features extracted by a convolutional neural network are mixed. The feature vector consisting of 298 features for classification learning are formed after a feature reduction process. In order to verify the performance of the proposed algorithm, experiments are carried out on the BraTS2017 dataset. The experimental results show that the proposed method can quickly detect and locate the tumor. The overall segmentation accuracy is improved distinctly with respect to 4 state-of-the-art approaches.
2021, 43(4): 1003-1013.
doi: 10.11999/JEIT200025
Abstract:
In order to improve the positioning accuracy of mobile robots in Simultaneous Localization And Mapping (SLAM), a bionic localization algorithm based on Growing Self-Organizing Map(GSOM) neural network is proposed. The method connects the activation characteristics of the place cells with the neural network output layer neurons to establish a response, and constructs a spatial topology map through the GSOM neural network, and uses the perceived distance information to realize the activation response of the place cells to estimate the position of the robot. The running path of the robot is restored in this way. The experimental results show that the cell spacing R has a great influence on the positioning accuracy. Choosing the appropriate cell spacing can effectively reduce the learning time of the neural network and improve the positioning accuracy. The average error of the algorithm is less than 0.153 m, and the positioning accuracy is 90.243%, which is better than the original algorithm. It is verified that the model established by the algorithm can realize the spatial position representation of the robot, improves the positioning accuracy of the object under the experimental scene, and shows good position estimation performance.
In order to improve the positioning accuracy of mobile robots in Simultaneous Localization And Mapping (SLAM), a bionic localization algorithm based on Growing Self-Organizing Map(GSOM) neural network is proposed. The method connects the activation characteristics of the place cells with the neural network output layer neurons to establish a response, and constructs a spatial topology map through the GSOM neural network, and uses the perceived distance information to realize the activation response of the place cells to estimate the position of the robot. The running path of the robot is restored in this way. The experimental results show that the cell spacing R has a great influence on the positioning accuracy. Choosing the appropriate cell spacing can effectively reduce the learning time of the neural network and improve the positioning accuracy. The average error of the algorithm is less than 0.153 m, and the positioning accuracy is 90.243%, which is better than the original algorithm. It is verified that the model established by the algorithm can realize the spatial position representation of the robot, improves the positioning accuracy of the object under the experimental scene, and shows good position estimation performance.
Continuous Seam Carving Algorithm Based on Just Noticeable Distortion Algorithm in Image Retargeting
2021, 43(4): 1014-1021.
doi: 10.11999/JEIT191050
Abstract:
Image retargeting technologies require important information preservation and less edge distortion during increasing/decreasing image size. The seam carving based algorithms, as the classic retargeting model, receive widespread attention in recent years. However, because of the discrete least energy seam searching strategy, the retargeting information can not be passed generation by generation, which causes retargeting distortions to prevail. To solve this problem, the Just Noticeable Distortion (JND) algorithm is proposed to detect the potential distribution of distortion information. Through the proposed energy weight Ew, the JND information can be passed to the following retargeting iteration for distortion reduction. According to the best knowledge, it is the first time to propose the seam carving algorithm in continuous way by the JND algorithm and energy weight, are the promising results also demonstrated compared with several new approaches at public database ‘Retarget Me’, qualitatively and quantitatively.
Image retargeting technologies require important information preservation and less edge distortion during increasing/decreasing image size. The seam carving based algorithms, as the classic retargeting model, receive widespread attention in recent years. However, because of the discrete least energy seam searching strategy, the retargeting information can not be passed generation by generation, which causes retargeting distortions to prevail. To solve this problem, the Just Noticeable Distortion (JND) algorithm is proposed to detect the potential distribution of distortion information. Through the proposed energy weight Ew, the JND information can be passed to the following retargeting iteration for distortion reduction. According to the best knowledge, it is the first time to propose the seam carving algorithm in continuous way by the JND algorithm and energy weight, are the promising results also demonstrated compared with several new approaches at public database ‘Retarget Me’, qualitatively and quantitatively.
2021, 43(4): 1022-1028.
doi: 10.11999/JEIT200778
Abstract:
The core of the dynamic flight pricing is to yield a pricing strategy with maximum seat revenue. The state-of-the-art flight pricing approaches are built on forecasting the fare demand. They suffer low profit due to the inaccurate prediction. To tackle the above issue, an approach for dynamic flight pricing based on strategy learning is proposed. That approach resorts to reinforcement learning to output pricing strategy with the highest expected return. That strategy is learned by iteratively policy evaluation and policy improvement. The rate of profit improvement on the two flights is empirically 30.94% and 39.96% over the existing pricing strategy, while that rate is 6.04% and 3.36% over the demand forecasting algorithm.
The core of the dynamic flight pricing is to yield a pricing strategy with maximum seat revenue. The state-of-the-art flight pricing approaches are built on forecasting the fare demand. They suffer low profit due to the inaccurate prediction. To tackle the above issue, an approach for dynamic flight pricing based on strategy learning is proposed. That approach resorts to reinforcement learning to output pricing strategy with the highest expected return. That strategy is learned by iteratively policy evaluation and policy improvement. The rate of profit improvement on the two flights is empirically 30.94% and 39.96% over the existing pricing strategy, while that rate is 6.04% and 3.36% over the demand forecasting algorithm.
2021, 43(4): 1029-1034.
doi: 10.11999/JEIT191009
Abstract:
Under the directed network topology, the consensus of the second order multi-agent system with large communication delay is studied, the protocol of delayed-state-derivative feedback with weighting term is proposed, which improves the problem of system oscillation caused by large communication delay. Firstly, the delayed-state-derivative feedback protocol with weighting term is introduced, and the closed-loop form of the second-order multi-agent system is given. Then, the sufficient and necessary conditions for the asymptotic steady-state consensus of the second-order multi-agent system are obtained by using the frequency domain analysis method, and it is proved that the second-order multi-agent system can tolerate the greater communication delay under the delayed-state-derivative feedback protocol with weighting term. Finally, the advantage of the delayed-state-derivative feedback protocol with weighting term is verified by numerical simulation.
Under the directed network topology, the consensus of the second order multi-agent system with large communication delay is studied, the protocol of delayed-state-derivative feedback with weighting term is proposed, which improves the problem of system oscillation caused by large communication delay. Firstly, the delayed-state-derivative feedback protocol with weighting term is introduced, and the closed-loop form of the second-order multi-agent system is given. Then, the sufficient and necessary conditions for the asymptotic steady-state consensus of the second-order multi-agent system are obtained by using the frequency domain analysis method, and it is proved that the second-order multi-agent system can tolerate the greater communication delay under the delayed-state-derivative feedback protocol with weighting term. Finally, the advantage of the delayed-state-derivative feedback protocol with weighting term is verified by numerical simulation.
2021, 43(4): 1035-1041.
doi: 10.11999/JEIT200026
Abstract:
In view of the nonlinear and chaos of urban short-term traffic flow sequence, this article proposes a combined prediction model based on multi-dimensional spatio-temporal Nonlinear Principal Component Analysis (NPCA) and Phase Space Reconstructed (PSR) Improved Gray Model (IGM(1,1)) in order to improve its forecast accuracy. First, the data correlation NPCA algorithm is used to reduce the spatial and temporal dimensions of multi-dimensional traffic flow sequences, while preserving the main traffic flow data that affects the predicted points, so as to improve the accuracy of the modeling. Phase space reconstruction amplifies the subtle features inside the traffic flow, so that its internal laws can be fully displayed, and improve further the prediction accuracy. Finally, the gray model combined with the improved background value is adapted to the characteristics of linearity, non-linearity and less required data. Short-term traffic flow is predicted. The experimental results show that the average relative error of the NPCA-PSR-IGM (1,1) combination prediction model is 3.12% smaller than that of the NPCA-PSR-GM (1,1) combination prediction model, and its standard deviation is relative to the PCA-PSR-IGM (1,1) combination prediction model has dropped from 15.7091 to 2.0589. At the same time, compared with the latest prediction model, the combined prediction model also improves the prediction accuracy and achieves a better prediction effect.
In view of the nonlinear and chaos of urban short-term traffic flow sequence, this article proposes a combined prediction model based on multi-dimensional spatio-temporal Nonlinear Principal Component Analysis (NPCA) and Phase Space Reconstructed (PSR) Improved Gray Model (IGM(1,1)) in order to improve its forecast accuracy. First, the data correlation NPCA algorithm is used to reduce the spatial and temporal dimensions of multi-dimensional traffic flow sequences, while preserving the main traffic flow data that affects the predicted points, so as to improve the accuracy of the modeling. Phase space reconstruction amplifies the subtle features inside the traffic flow, so that its internal laws can be fully displayed, and improve further the prediction accuracy. Finally, the gray model combined with the improved background value is adapted to the characteristics of linearity, non-linearity and less required data. Short-term traffic flow is predicted. The experimental results show that the average relative error of the NPCA-PSR-IGM (1,1) combination prediction model is 3.12% smaller than that of the NPCA-PSR-GM (1,1) combination prediction model, and its standard deviation is relative to the PCA-PSR-IGM (1,1) combination prediction model has dropped from 15.7091 to 2.0589. At the same time, compared with the latest prediction model, the combined prediction model also improves the prediction accuracy and achieves a better prediction effect.
2021, 43(4): 1042-1049.
doi: 10.11999/JEIT200353
Abstract:
Resistance spot welding is a complex process in which many factors interact. Given the small size of data sets available and the complex nature of unstable processes, it is difficult to establish an accurate mathematical model to predict the parameters of resistance spot welding. An optimal computing method for solving this problem is presented. The method combines Bayes-XGBoost with the Particle Swarm Optimization (PSO) algorithm to select suitable features and to enable the optimal combinations of samples for 0.15 mm nickel sheets and for 0.4 mm stainless steel battery positive caps; The non-linear slicing ability and anti-overfitting mechanism of eXtreme Gradient Boosting (XGBoost) are used to train forward spot welding parameters; and Bayesian optimization is applied to the XGBoost's optimal parameter selection. The method uses the global optimization feature of Particle Swarm Optimization (PSO) to predict the backward process parameters with variable target values such that the optimal process parameters are obtained. Compared with other algorithms mentioned in this paper, this method offers more comprehensive performance and possesses better capabilities to effectively assist in the spot welding process, which are demonstrated by the resistance spot welding experiments performed.
Resistance spot welding is a complex process in which many factors interact. Given the small size of data sets available and the complex nature of unstable processes, it is difficult to establish an accurate mathematical model to predict the parameters of resistance spot welding. An optimal computing method for solving this problem is presented. The method combines Bayes-XGBoost with the Particle Swarm Optimization (PSO) algorithm to select suitable features and to enable the optimal combinations of samples for 0.15 mm nickel sheets and for 0.4 mm stainless steel battery positive caps; The non-linear slicing ability and anti-overfitting mechanism of eXtreme Gradient Boosting (XGBoost) are used to train forward spot welding parameters; and Bayesian optimization is applied to the XGBoost's optimal parameter selection. The method uses the global optimization feature of Particle Swarm Optimization (PSO) to predict the backward process parameters with variable target values such that the optimal process parameters are obtained. Compared with other algorithms mentioned in this paper, this method offers more comprehensive performance and possesses better capabilities to effectively assist in the spot welding process, which are demonstrated by the resistance spot welding experiments performed.
2021, 43(4): 1050-1056.
doi: 10.11999/JEIT200091
Abstract:
With the continuous development of the Internet of Things(IoT), its business demands show a trend of diversification and globalization. As the ground Internet of Things can not cover the whole world, the satellite IoT, especially the Low Earth Orbit Satellite Constellation (LEOSC) IoT, can supplement and extend the ground network. Due to the wide coverage and high dynamic characteristics of the LEOSC IoT system, there are significant differences between it and the ground IoT in terms of traffic statistics. In order to make reasonable and efficient use of limited resources on board, the traffic model of global Internet of Things based on LEOSC is studied in this paper. Combined with diversified traffic characteristics and satellite communication system characteristics, the framework of global IoT traffic model is obtained by using statistical modeling theory. What’s more, an access strategy based on the highest priority is proposed to enable the device node to select the satellite in real time. The simulation results show that the Poisson process can be used to simulate approximately the superposition process of asynchronous traffic commonly exist in LEOSC IoT, and due to the high dynamic nature of low earth orbit satellite, its traffic source changes at high speed, resulting in high Peak-to-Average Ratio(PAR) of traffic.
With the continuous development of the Internet of Things(IoT), its business demands show a trend of diversification and globalization. As the ground Internet of Things can not cover the whole world, the satellite IoT, especially the Low Earth Orbit Satellite Constellation (LEOSC) IoT, can supplement and extend the ground network. Due to the wide coverage and high dynamic characteristics of the LEOSC IoT system, there are significant differences between it and the ground IoT in terms of traffic statistics. In order to make reasonable and efficient use of limited resources on board, the traffic model of global Internet of Things based on LEOSC is studied in this paper. Combined with diversified traffic characteristics and satellite communication system characteristics, the framework of global IoT traffic model is obtained by using statistical modeling theory. What’s more, an access strategy based on the highest priority is proposed to enable the device node to select the satellite in real time. The simulation results show that the Poisson process can be used to simulate approximately the superposition process of asynchronous traffic commonly exist in LEOSC IoT, and due to the high dynamic nature of low earth orbit satellite, its traffic source changes at high speed, resulting in high Peak-to-Average Ratio(PAR) of traffic.
2021, 43(4): 1057-1063.
doi: 10.11999/JEIT200141
Abstract:
In order to overcome the complicated implementation to process tail data in high bit-width Cyclic Redundancy Check(CRC) calculation for variable length packet, linear matrix computation is used to investigate CRC inverse calculation. And a rollback algorithm is introduced to simplify the regular algorithm. Then the experiment is conducted to implement the rollback algorithm in Altera FPGA device. The results show that rollback algorithm utilizes fewer resource and is more easily to implement. In 512 bit data width variable length CRC calculation implement in FPGA, the resource utilization is decreased to 15% of regular algorithm by applying rollback algorithm. Synthesis time is decreased to 30%, and Place&Route time is deceased to 40%. It is concluded that the new rollback algorithm has great advantage.
In order to overcome the complicated implementation to process tail data in high bit-width Cyclic Redundancy Check(CRC) calculation for variable length packet, linear matrix computation is used to investigate CRC inverse calculation. And a rollback algorithm is introduced to simplify the regular algorithm. Then the experiment is conducted to implement the rollback algorithm in Altera FPGA device. The results show that rollback algorithm utilizes fewer resource and is more easily to implement. In 512 bit data width variable length CRC calculation implement in FPGA, the resource utilization is decreased to 15% of regular algorithm by applying rollback algorithm. Synthesis time is decreased to 30%, and Place&Route time is deceased to 40%. It is concluded that the new rollback algorithm has great advantage.
2021, 43(4): 1064-1071.
doi: 10.11999/JEIT191049
Abstract:
In view of the challenges of 5G communication technology with high speed and multiple service scenarios, this paper proposes a new component-based Software Defined wireless access Network (SDN) architecture. Based on the architecture of Centralized Unit (CU), Distributed Unit (DU) and Active Antenna Unit(AAU) in 5G access network, further component-based evolution is carried out to form a new architecture composed the communication units of Centralized Control Unit (CCU), CU, DU, Radio Unit(RU), and AAU. This new architecture is not only conducive to the realization of wireless access network with slicing and virtualization, but also conducive to the adoption of distributed computing technology and hardware accelerating technology to break through the processing bottleneck of general-purpose processor, and reduce the forward transmission pressure between DU and AAU. In this paper, a prototype of component-based soft base station is developed and tested. The results show that the component-based scheme can not only provide high flexibility, but also improve the processing capacity of the general processor soft base station and reduce effectively the traffic of remote stations.
In view of the challenges of 5G communication technology with high speed and multiple service scenarios, this paper proposes a new component-based Software Defined wireless access Network (SDN) architecture. Based on the architecture of Centralized Unit (CU), Distributed Unit (DU) and Active Antenna Unit(AAU) in 5G access network, further component-based evolution is carried out to form a new architecture composed the communication units of Centralized Control Unit (CCU), CU, DU, Radio Unit(RU), and AAU. This new architecture is not only conducive to the realization of wireless access network with slicing and virtualization, but also conducive to the adoption of distributed computing technology and hardware accelerating technology to break through the processing bottleneck of general-purpose processor, and reduce the forward transmission pressure between DU and AAU. In this paper, a prototype of component-based soft base station is developed and tested. The results show that the component-based scheme can not only provide high flexibility, but also improve the processing capacity of the general processor soft base station and reduce effectively the traffic of remote stations.
2021, 43(4): 1072-1079.
doi: 10.11999/JEIT200017
Abstract:
With the rapid development of the Internet of Vehicles (IoV), the number of cars and users requesting tasks offloading is also increasing. The Mobile Edge Computing (MEC) can effectively solve the challenge of high offload transmission delays for task offloading in communication network, but there still is a problem that the channel resources are insufficient in the network model. Compared with traditional Orthogonal Multiple Access (OMA), the technology of Non-Orthogonal Multiple Access (NOMA) can service more users with task offload under the same channel resource conditions. In this paper, considering the multiple aspects of task offloading impact factor, a mixed unloading strategy based on NOMA-MEC is proposed. A game algorithm based on Deep Q-learning Network (DQN) is designed to make channel selection for vehicle users and provide an optimal power allocation strategy through multiple iterative learning of neural networks. The simulation results show that the proposed hybrid NOMA-MEC offloading strategy can effectively optimize the multi-user offloading delay and energy consumption and ensure maximize the benefits of users.
With the rapid development of the Internet of Vehicles (IoV), the number of cars and users requesting tasks offloading is also increasing. The Mobile Edge Computing (MEC) can effectively solve the challenge of high offload transmission delays for task offloading in communication network, but there still is a problem that the channel resources are insufficient in the network model. Compared with traditional Orthogonal Multiple Access (OMA), the technology of Non-Orthogonal Multiple Access (NOMA) can service more users with task offload under the same channel resource conditions. In this paper, considering the multiple aspects of task offloading impact factor, a mixed unloading strategy based on NOMA-MEC is proposed. A game algorithm based on Deep Q-learning Network (DQN) is designed to make channel selection for vehicle users and provide an optimal power allocation strategy through multiple iterative learning of neural networks. The simulation results show that the proposed hybrid NOMA-MEC offloading strategy can effectively optimize the multi-user offloading delay and energy consumption and ensure maximize the benefits of users.
2021, 43(4): 1080-1089.
doi: 10.11999/JEIT200269
Abstract:
The performance of Generalized Frequency Division Multiplexing (GFDM) systems significantly degrades over time-frequency doubly selective channels due to the severe inter-carrier interference and inter-subsymbol interference. To this end, a Time-Offset GFDM (TO-GFDM) is proposed, which can improve the performance of GFDM systems under doubly selective channel environment by introducing a time-offset to the prototype filter of conventional GFDM systems. The average signal-to-interference ratio of GFDM systems in doubly selective channels is analytically derived, and a pilot-aided joint iterative channel estimation and symbol detection algorithm is proposed. The proposed algorithm can progressively decrease the interference signal and improve the performance of channel estimation and symbol detection with the information exchanging between the channel estimation unit and the successive interference cancellation based symbol detection unit. The results of theoretical analysis and simulation demonstrate that in doubly selective channels, the time-offset GFDM outperforms the conventional GFDM in terms of average signal-to-interference ratio and bit error rate, and the joint iterative channel estimation and symbol detection algorithm can reduce the bit error rate effectively.
The performance of Generalized Frequency Division Multiplexing (GFDM) systems significantly degrades over time-frequency doubly selective channels due to the severe inter-carrier interference and inter-subsymbol interference. To this end, a Time-Offset GFDM (TO-GFDM) is proposed, which can improve the performance of GFDM systems under doubly selective channel environment by introducing a time-offset to the prototype filter of conventional GFDM systems. The average signal-to-interference ratio of GFDM systems in doubly selective channels is analytically derived, and a pilot-aided joint iterative channel estimation and symbol detection algorithm is proposed. The proposed algorithm can progressively decrease the interference signal and improve the performance of channel estimation and symbol detection with the information exchanging between the channel estimation unit and the successive interference cancellation based symbol detection unit. The results of theoretical analysis and simulation demonstrate that in doubly selective channels, the time-offset GFDM outperforms the conventional GFDM in terms of average signal-to-interference ratio and bit error rate, and the joint iterative channel estimation and symbol detection algorithm can reduce the bit error rate effectively.
2021, 43(4): 1090-1097.
doi: 10.11999/JEIT190785
Abstract:
Taking the advantage of the parallel development of electronic sampling systems and signal processing, Digital Radio-over-Fiber (DRoF) is studied extensively as a way of providing multi-service transmission at low-bit rate through data compression. However, the dynamic range is greatly lowered after data compression in the system. Based on the theoretical analysis of the compression parameters, a novel Fast-Settling Two-stage Automatic Gain Control (FST-AGC) algorithm is proposed, in which gain adjustment settling is carried out by multi-threshold decision mechanism with a fast-settling times, high stability and great accuracy. By introducing a novel gain control mechanism which simultaneously adjusts the gain in digital domain and Radio Frequency (RF) domain, the dynamic range of the system increases significantly. This algorithm has been applied to a DRoF system which supports the low-bit rate transmission of all services of 3 China Mobile Network Operators (MONs) successfully. The theoretical analysis, the simulation results and the experimental data all prove the validity of the proposed algorithm. Its promising properties and excellent performance enable its potential application to emerging networks, such as Internet of Things (IoT), Radio Frequency Intification (RFID) and the incoming 5G network.
Taking the advantage of the parallel development of electronic sampling systems and signal processing, Digital Radio-over-Fiber (DRoF) is studied extensively as a way of providing multi-service transmission at low-bit rate through data compression. However, the dynamic range is greatly lowered after data compression in the system. Based on the theoretical analysis of the compression parameters, a novel Fast-Settling Two-stage Automatic Gain Control (FST-AGC) algorithm is proposed, in which gain adjustment settling is carried out by multi-threshold decision mechanism with a fast-settling times, high stability and great accuracy. By introducing a novel gain control mechanism which simultaneously adjusts the gain in digital domain and Radio Frequency (RF) domain, the dynamic range of the system increases significantly. This algorithm has been applied to a DRoF system which supports the low-bit rate transmission of all services of 3 China Mobile Network Operators (MONs) successfully. The theoretical analysis, the simulation results and the experimental data all prove the validity of the proposed algorithm. Its promising properties and excellent performance enable its potential application to emerging networks, such as Internet of Things (IoT), Radio Frequency Intification (RFID) and the incoming 5G network.
2021, 43(4): 1098-1105.
doi: 10.11999/JEIT200598
Abstract:
In the process of solving the problem of slow impedance calculation speed in the design of Radio Frequency IDentification (RFID) tag antenna, a method of impedance prediction based on polynomial for folded dipole RFID tag antenna is proposed in view of the complex impedance coupling. Firstly, impedance transformation based on antenna size and linearization assumption are used to establish model hypothesis. Then the data are collected from the antenna structure and the correlation analysis and regression fitting are carried out to verify the validity of the hypothesis. Finally, the accuracy, efficiency and universality of the impedance prediction method compared with computer simulation are verified by experiments. The experimental results show that the predicted impedance can greatly shorten the calculation time while maintaining high prediction accuracy when the proposed method replace computer to calculate the impedance of the folded dipole RFID tag antenna, and the method is still applicable to the RFID tag antenna with different bending times in the frequency band used in China.
In the process of solving the problem of slow impedance calculation speed in the design of Radio Frequency IDentification (RFID) tag antenna, a method of impedance prediction based on polynomial for folded dipole RFID tag antenna is proposed in view of the complex impedance coupling. Firstly, impedance transformation based on antenna size and linearization assumption are used to establish model hypothesis. Then the data are collected from the antenna structure and the correlation analysis and regression fitting are carried out to verify the validity of the hypothesis. Finally, the accuracy, efficiency and universality of the impedance prediction method compared with computer simulation are verified by experiments. The experimental results show that the predicted impedance can greatly shorten the calculation time while maintaining high prediction accuracy when the proposed method replace computer to calculate the impedance of the folded dipole RFID tag antenna, and the method is still applicable to the RFID tag antenna with different bending times in the frequency band used in China.
2021, 43(4): 1106-1111.
doi: 10.11999/JEIT200277
Abstract:
Compared with the traditional continuous model, the mixed continuous model weakens the impedance condition and simplifies the difficulty of broadband matching. In this paper, a new type of harmonic control network used a hybrid continuous model and based on the concept of impedance buffering is designed to design a hybrid continuous radio frequency power amplifier that spans three octave layers. The measured results show that the drain efficiency is 58.4%～72.6%, the gain is more than 10 dB, and the output power is 39.8～41.2 dBm in the frequency range of 168.4% relative bandwidth of 0.3～3.5 GHz.
Compared with the traditional continuous model, the mixed continuous model weakens the impedance condition and simplifies the difficulty of broadband matching. In this paper, a new type of harmonic control network used a hybrid continuous model and based on the concept of impedance buffering is designed to design a hybrid continuous radio frequency power amplifier that spans three octave layers. The measured results show that the drain efficiency is 58.4%～72.6%, the gain is more than 10 dB, and the output power is 39.8～41.2 dBm in the frequency range of 168.4% relative bandwidth of 0.3～3.5 GHz.
2021, 43(4): 1112-1119.
doi: 10.11999/JEIT200046
Abstract:
This paper focuses on the Low-Density-Parity-Check (LDPC) decoder for 5G New Radio (NR) specification. After introducing the characteristics of the LDPC code in 5G NR, the performance of different decoding algorithms are compared, and then the overall architecture of the decoder is proposed. In the proposed architecture, the decoder is divided into high-speed decoder and high-performance decoder. The high-speed decoder is intended for high-rate and high throughput decoding, while the high-performance decoder is used for low-rate decoding under low Signal-to-Noise-Ratio (SNR) scenarios, which is for communications under extremely bad situations, and does not need a high throughput. The design is implemented on Field Programmable Gate Array (FPGA) and the results are shown.
This paper focuses on the Low-Density-Parity-Check (LDPC) decoder for 5G New Radio (NR) specification. After introducing the characteristics of the LDPC code in 5G NR, the performance of different decoding algorithms are compared, and then the overall architecture of the decoder is proposed. In the proposed architecture, the decoder is divided into high-speed decoder and high-performance decoder. The high-speed decoder is intended for high-rate and high throughput decoding, while the high-performance decoder is used for low-rate decoding under low Signal-to-Noise-Ratio (SNR) scenarios, which is for communications under extremely bad situations, and does not need a high throughput. The design is implemented on Field Programmable Gate Array (FPGA) and the results are shown.
2021, 43(4): 1120-1128.
doi: 10.11999/JEIT200068
Abstract:
The major drawbacks of MultiCarrier Differential Chaos Shift Keying (MC-DCSK) system are relating to low data rate and poor bit error performance. Therefore a Quadrature MultiCarrier Noise Reduction Differential Chaos Shift Keying (QMC-NR-DCSK) system is proposed to improve the performances of MC-DCSK system. At the transmitter, reference signal is transmitted on the predefined carrier. While the remaining M-1 carriers and the carriers with orthogonal phase at the same frequency are all used to transmit information signals by using Quadrature Modulation technology, and the data-rate to bandwidth ratio and transmission rate of which are four times higher than that of MC-DCSK system by further introducing Hilbert transformation. The noise reduction operation of the moving average filter is introduced to reduce the variance of noise at the receiver, thereby improving the bit error performance of the QMC-NR-DCSK system. The bit error rate formula and simulations of QMC-NR-DCSK system under Additive White Gaussian Noise (AWGN) channel and multi-path Rayleigh Fading Channel (RFC) are carried out respectively. The results show that QMC-NR-DCSK system can effectively improve the transmission rate, data-rate to bandwidth ratio and bit error performance, which provides a theoretical reference for the application of the multicarrier communication systems.
The major drawbacks of MultiCarrier Differential Chaos Shift Keying (MC-DCSK) system are relating to low data rate and poor bit error performance. Therefore a Quadrature MultiCarrier Noise Reduction Differential Chaos Shift Keying (QMC-NR-DCSK) system is proposed to improve the performances of MC-DCSK system. At the transmitter, reference signal is transmitted on the predefined carrier. While the remaining M-1 carriers and the carriers with orthogonal phase at the same frequency are all used to transmit information signals by using Quadrature Modulation technology, and the data-rate to bandwidth ratio and transmission rate of which are four times higher than that of MC-DCSK system by further introducing Hilbert transformation. The noise reduction operation of the moving average filter is introduced to reduce the variance of noise at the receiver, thereby improving the bit error performance of the QMC-NR-DCSK system. The bit error rate formula and simulations of QMC-NR-DCSK system under Additive White Gaussian Noise (AWGN) channel and multi-path Rayleigh Fading Channel (RFC) are carried out respectively. The results show that QMC-NR-DCSK system can effectively improve the transmission rate, data-rate to bandwidth ratio and bit error performance, which provides a theoretical reference for the application of the multicarrier communication systems.
2021, 43(4): 1129-1136.
doi: 10.11999/JEIT200032
Abstract:
Resource allocation in Multi-Cell hybrid Non-Orthogonal Multiple Access-orthogonal multiple access (MC-hybrid NOMA) networks is studied in this paper. To satisfy the Quality of Experience (QoE) of different service types of users, an algorithm joint user-BS association, sub-channel assignment and power allocation is proposed to maximize the sum Mean Opinion Scores (MOSs) of users in the networks. A low-complexity two-step approach based on matching game theory and developed power allocation strategy based on QoE proportional fairness are proposed. Simulation results demonstrate that the proposed algorithm can effectively improve the system performance and fairness.
Resource allocation in Multi-Cell hybrid Non-Orthogonal Multiple Access-orthogonal multiple access (MC-hybrid NOMA) networks is studied in this paper. To satisfy the Quality of Experience (QoE) of different service types of users, an algorithm joint user-BS association, sub-channel assignment and power allocation is proposed to maximize the sum Mean Opinion Scores (MOSs) of users in the networks. A low-complexity two-step approach based on matching game theory and developed power allocation strategy based on QoE proportional fairness are proposed. Simulation results demonstrate that the proposed algorithm can effectively improve the system performance and fairness.
2021, 43(4): 1137-1144.
doi: 10.11999/JEIT200019
Abstract:
For the problem of long code Direct Sequence-Code Division Multiple Access (DS-CDMA) signal in traditional asynchronous single-channel with multipath effect under low Signal-to-Noise Ratio (SNR), including blind estimation of the Pseudo-Noise (PN) sequence and information sequence, a method using multi-channel synchronous and asynchronous based on parallel factor is proposed. Firstly, the received signal in multipath environment is modeled as a multi-channel receiving model. And then the long code DS-CDMA signal is modeled as the short code DS-CDMA signal with missing data to form the observation missing-data matrix, which is equivalent to a parallel factor model with missing data. Finally, the regularized Alternating Least Squares (ALS) algorithm is applied to decompose the parallel factor, and the PN sequence and information sequence of long code DS-CDMA signals in multipath environment can be further estimated. Simulation results show that the performance of sequences estimation closely relates with the multipath environment, and the estimation error rate of 6 user PN sequences and information sequence is less than 1% under the condition that the Rician factor is equal to 10 and the number of path and channel are 3 and 4 respectively when the SNR is higher than -10 dB.
For the problem of long code Direct Sequence-Code Division Multiple Access (DS-CDMA) signal in traditional asynchronous single-channel with multipath effect under low Signal-to-Noise Ratio (SNR), including blind estimation of the Pseudo-Noise (PN) sequence and information sequence, a method using multi-channel synchronous and asynchronous based on parallel factor is proposed. Firstly, the received signal in multipath environment is modeled as a multi-channel receiving model. And then the long code DS-CDMA signal is modeled as the short code DS-CDMA signal with missing data to form the observation missing-data matrix, which is equivalent to a parallel factor model with missing data. Finally, the regularized Alternating Least Squares (ALS) algorithm is applied to decompose the parallel factor, and the PN sequence and information sequence of long code DS-CDMA signals in multipath environment can be further estimated. Simulation results show that the performance of sequences estimation closely relates with the multipath environment, and the estimation error rate of 6 user PN sequences and information sequence is less than 1% under the condition that the Rician factor is equal to 10 and the number of path and channel are 3 and 4 respectively when the SNR is higher than -10 dB.
Stagger Pulse Repetition Interval Pulse Train Deinterleaving Algorithm Based on Sequence Association
2021, 43(4): 1145-1153.
doi: 10.11999/JEIT191030
Abstract:
For the conventional histogram signal deinterleaving algorithm’s drawback of stagger signal, a histogram algorithm based on corresponding of pulse interval and single pulse is proposed. This algorithm utilizes the corresponding of pulse pair interval and single pulse to get a matrix named Pulse Interval Distribution Matrix (PIDM), and a novel histogram is obtained via cumulating row of the matrix. This histogram can avoid the suppressing of frame period of staggered pulse train caused by Pulse Repetition Intervals (PRI) transform algorithm when staggering signal deinterleaving, and can extract the subsequence of pulse train through PIDM. Simulation results show the algorithm has excellent performance on environment including multi staggered pulse trains with multi-fixed pulse trains under the circumstance of without adding the complexity of calculating.
For the conventional histogram signal deinterleaving algorithm’s drawback of stagger signal, a histogram algorithm based on corresponding of pulse interval and single pulse is proposed. This algorithm utilizes the corresponding of pulse pair interval and single pulse to get a matrix named Pulse Interval Distribution Matrix (PIDM), and a novel histogram is obtained via cumulating row of the matrix. This histogram can avoid the suppressing of frame period of staggered pulse train caused by Pulse Repetition Intervals (PRI) transform algorithm when staggering signal deinterleaving, and can extract the subsequence of pulse train through PIDM. Simulation results show the algorithm has excellent performance on environment including multi staggered pulse trains with multi-fixed pulse trains under the circumstance of without adding the complexity of calculating.
2021, 43(4): 1154-1161.
doi: 10.11999/JEIT191024
Abstract:
Passive radar uses third party radiation source, which is uncontrollable. Due to the complicated electromagnetic propagation conditions, especially in a low altitude target detection, the detection performance of the radar is greatly affected by clutters, leading to significant degradation of the performance of traditional constant false alarm algorithm. In order to improve the detection performance, a Dynamic Ordered Matrix Constant False Alarm Rate (DOM-CFAR) algorithm based on radar clutter space partition is proposed. In this algorithm, an ordered matrix is constructed after dividing the clutter space from distance and Doppler dimension. Then the dynamic extreme value is replaced according to the background clutter change and the estimated median value of clutter is extracted so as to calculate the detection threshold. This algorithm makes the threshold value of detection can dynamically adapt to the clutter power change. The simulation and measurement results show that the algorithm can maintain excellent detection performance under the complex environment with uniform clutter, multi-target and clutter edge.
Passive radar uses third party radiation source, which is uncontrollable. Due to the complicated electromagnetic propagation conditions, especially in a low altitude target detection, the detection performance of the radar is greatly affected by clutters, leading to significant degradation of the performance of traditional constant false alarm algorithm. In order to improve the detection performance, a Dynamic Ordered Matrix Constant False Alarm Rate (DOM-CFAR) algorithm based on radar clutter space partition is proposed. In this algorithm, an ordered matrix is constructed after dividing the clutter space from distance and Doppler dimension. Then the dynamic extreme value is replaced according to the background clutter change and the estimated median value of clutter is extracted so as to calculate the detection threshold. This algorithm makes the threshold value of detection can dynamically adapt to the clutter power change. The simulation and measurement results show that the algorithm can maintain excellent detection performance under the complex environment with uniform clutter, multi-target and clutter edge.
2021, 43(4): 1162-1169.
doi: 10.11999/JEIT200042
Abstract:
Single-observer Multi-illuminator radars localize multiple moving targets using multiple Bistatic Range (BR) and Bistatic Range Rate (BRR) measurements. However the existence of the biases would degrade the localization performance. A novel joint localization and registration based on iterative post-estimation Dependent Least Squares (DLS) is proposed. Firstly, a new jointed target parameters and biases estimation model is given by linearizing the BR and BRR measurement equation with auxiliary parameters. Then a DLS algorithm refines the target parameters by reusing the relationships between the target parameters and auxiliary parameters. An iterative post-estimate further refines the estimation accuracy and global coverage by biases registration. Finally its theoretical error and global convergence are obtained analytically. Simulation results show that the proposed algorithm has a good global convergence behavior and the performance can achieve the Cramer-Rao Lower Bound.
Single-observer Multi-illuminator radars localize multiple moving targets using multiple Bistatic Range (BR) and Bistatic Range Rate (BRR) measurements. However the existence of the biases would degrade the localization performance. A novel joint localization and registration based on iterative post-estimation Dependent Least Squares (DLS) is proposed. Firstly, a new jointed target parameters and biases estimation model is given by linearizing the BR and BRR measurement equation with auxiliary parameters. Then a DLS algorithm refines the target parameters by reusing the relationships between the target parameters and auxiliary parameters. An iterative post-estimate further refines the estimation accuracy and global coverage by biases registration. Finally its theoretical error and global convergence are obtained analytically. Simulation results show that the proposed algorithm has a good global convergence behavior and the performance can achieve the Cramer-Rao Lower Bound.
2021, 43(4): 1170-1176.
doi: 10.11999/JEIT191061
Abstract:
Integration has become the most important development direction of modern electronic system, and so is radar system. Modern direct-conversion technology has the characteristics of low cost, small size, simple structure and high integration. It has been widely used in the communication market, such as mobile phones, base stations, satellite receivers and GPS receivers, etc. But up to now, there has been no report of practical application to radar. This paper studies the application of direct-conversion technology to radar, especially to phased array radar, which can greatly reduce the cost and volume of radar, and has the characteristics of wide working frequency band, flexible configuration and reconfiguration. Therefore, direct-conversion technology will also become a development direction of radar technology. Finally, a design example of applying direct-conversion technology to S band phased array radar is given.
Integration has become the most important development direction of modern electronic system, and so is radar system. Modern direct-conversion technology has the characteristics of low cost, small size, simple structure and high integration. It has been widely used in the communication market, such as mobile phones, base stations, satellite receivers and GPS receivers, etc. But up to now, there has been no report of practical application to radar. This paper studies the application of direct-conversion technology to radar, especially to phased array radar, which can greatly reduce the cost and volume of radar, and has the characteristics of wide working frequency band, flexible configuration and reconfiguration. Therefore, direct-conversion technology will also become a development direction of radar technology. Finally, a design example of applying direct-conversion technology to S band phased array radar is given.
2021, 43(4): 1177-1184.
doi: 10.11999/JEIT190999
Abstract:
Rotating phased array multi-function radar have electronically scanning capability both in elevation and in azimuth. It can allocate time and energy both in azimuth and elevation more flexible than the mechanical scanning radar. In order to optimize the effects of search, it is needed to divide and assess the areas. Based on the calculation of the most threatening path of the rotating phased array radar, this paper presents a novel method of regional threat level assessment; The radar problem of most threatening trajectories computation may be declined to a shortest path problem solved by calculus of variations by taking the detection probability as the price function; The shortest path computation based on Eikonal equations are solved with the fast marching method. The most threatening path are computed by backtracking the minimal path based on steepest gradient descent method. Finally, the solution of rotating phased array multi-function radar resources management is proposed based on divided areas, and the validity of the technique is checked and rationality of constructed model is verified by providing instances with 240 typical targets.
Rotating phased array multi-function radar have electronically scanning capability both in elevation and in azimuth. It can allocate time and energy both in azimuth and elevation more flexible than the mechanical scanning radar. In order to optimize the effects of search, it is needed to divide and assess the areas. Based on the calculation of the most threatening path of the rotating phased array radar, this paper presents a novel method of regional threat level assessment; The radar problem of most threatening trajectories computation may be declined to a shortest path problem solved by calculus of variations by taking the detection probability as the price function; The shortest path computation based on Eikonal equations are solved with the fast marching method. The most threatening path are computed by backtracking the minimal path based on steepest gradient descent method. Finally, the solution of rotating phased array multi-function radar resources management is proposed based on divided areas, and the validity of the technique is checked and rationality of constructed model is verified by providing instances with 240 typical targets.
2021, 43(4): 1185-1191.
doi: 10.11999/JEIT200071
Abstract:
The three-Dimensional (3D) radar imaging mathods based on sparse representation by the scattering intensity of imaging sceen has a poor representation of geometric details on the shape of the target, which isn’t conducive to target recognition. Firstly, the structural characteristics of scattering intensity in the imaging scenario are analyzed in this paper. Then, by the structured sparse representation with the gradient information of scattering points, a structured sparse reconstruction model is constructed. Finally, the 3D imaging result is reconstructed by a improved joint Orthogonal Matching Pursuit (OMP) algorithm. The experimental results show that the proposed method has good anti-noise and imaging quality, and can reflect the geometric details of the target.
The three-Dimensional (3D) radar imaging mathods based on sparse representation by the scattering intensity of imaging sceen has a poor representation of geometric details on the shape of the target, which isn’t conducive to target recognition. Firstly, the structural characteristics of scattering intensity in the imaging scenario are analyzed in this paper. Then, by the structured sparse representation with the gradient information of scattering points, a structured sparse reconstruction model is constructed. Finally, the 3D imaging result is reconstructed by a improved joint Orthogonal Matching Pursuit (OMP) algorithm. The experimental results show that the proposed method has good anti-noise and imaging quality, and can reflect the geometric details of the target.
2021, 43(4): 1192-1198.
doi: 10.11999/JEIT200078
Abstract:
An optimal path planning problem is investigated for Angle-Of-Arrival (AOA) source localization using Unmanned Aerial Vehicles (UAVs) equipped with passive sensors. The more realistic model is considered where the variance of AOA measurement noises is a function of the source-to-sensor distances, which complicates AOA-based source localization. A modified Variable Gain Unscented Kalman Filter (VG-UKF)is developed to adapt to distance-dependent variance of AOA measurement noises. The Generalized Cramer-Rao Lower Bound (GCRLB) of AOA localization is calculated. Further, the unconstrained optimal sensor placement and constrained optimal sensor placement are theoretically analyzed. Then a path planning model for UAVs is constructed with minimizing the trace of GCRLB, which is solved optimally with penalty function and LM (Levenberg-Marquardt) algorithm. The effectiveness of the proposed algorithm is illustrated with simulation results.
An optimal path planning problem is investigated for Angle-Of-Arrival (AOA) source localization using Unmanned Aerial Vehicles (UAVs) equipped with passive sensors. The more realistic model is considered where the variance of AOA measurement noises is a function of the source-to-sensor distances, which complicates AOA-based source localization. A modified Variable Gain Unscented Kalman Filter (VG-UKF)is developed to adapt to distance-dependent variance of AOA measurement noises. The Generalized Cramer-Rao Lower Bound (GCRLB) of AOA localization is calculated. Further, the unconstrained optimal sensor placement and constrained optimal sensor placement are theoretically analyzed. Then a path planning model for UAVs is constructed with minimizing the trace of GCRLB, which is solved optimally with penalty function and LM (Levenberg-Marquardt) algorithm. The effectiveness of the proposed algorithm is illustrated with simulation results.
