Email alert
2023 Vol. 45, No. 12
Display Method:
2023, 45(12): 4203-4210.
doi: 10.11999/JEIT221189
Abstract:
In this paper, secure transmission in Intelligent Reflecting Surface (IRS) assisted Unmanned Aerial Vehicle (UAV) networks based on Non-Orthogonal Multiple Access (NOMA) is investigated. A joint placement and transmit power of UAV, successive interference cancellation decoding orders, and reflecting matrix of IRS optimization problem is formulated to maximize the secrecy rate. Since the problem is mixed-integer non-convex and challenging to solve, a block coordinate descent based iterative algorithm is developed. The original problem is decomposed into three subproblems, which are solved by exploiting the penalty-based method, the semidefinite relaxation technique, and the successive convex approximation technique. Simulation results demonstrate that the security rate of the proposed scheme is better than the scheme without IRS assisted NOMA network and the scheme without IRS assisted orthogonal multiple access network.
In this paper, secure transmission in Intelligent Reflecting Surface (IRS) assisted Unmanned Aerial Vehicle (UAV) networks based on Non-Orthogonal Multiple Access (NOMA) is investigated. A joint placement and transmit power of UAV, successive interference cancellation decoding orders, and reflecting matrix of IRS optimization problem is formulated to maximize the secrecy rate. Since the problem is mixed-integer non-convex and challenging to solve, a block coordinate descent based iterative algorithm is developed. The original problem is decomposed into three subproblems, which are solved by exploiting the penalty-based method, the semidefinite relaxation technique, and the successive convex approximation technique. Simulation results demonstrate that the security rate of the proposed scheme is better than the scheme without IRS assisted NOMA network and the scheme without IRS assisted orthogonal multiple access network.
2023, 45(12): 4211-4217.
doi: 10.11999/JEIT221289
Abstract:
In this paper, a dual-channel nonlinear feedback architecture is proposed to suppress nonlinear distortion of Power Amplifiers (PA) in the analog domain to improve PA linearity and reduce adjacent channel leakage. In this architecture, a nonlinear extraction loop and a feedback adjustment loop are included to suppress the nonlinearity. First, in the nonlinear extraction loop, the PA input and output signals are extracted by a coupler and aligned with amplitude and phase. Then the two signals are combined to cancel the linear signal and obtain the nonlinear distortion generated by PA. Next, in the feedback adjustment loop, two independent analog channels are used to modify the amplitude and delay of the extracted nonlinear signal before injecting into the PA input port. The delays of these two channels can be finely tuned to ensure the whole feedback structure to behave like a second-order Delta-Sigma modulator, and it shows better distortion suppression performance compared with the single-channel nonlinear feedback architecture. By configuring the feedback channel parameters through the proposed method, flexible suppression of nonlinear distortion at different target frequency ranges can be achieved. An experimental platform using a commercial PA chip CMPA0060002F is designed for verification. For a test signal with a bandwidth of 40 MHz and a carrier frequency of 780 MHz, under the current hardware feedback delay of 6 ns, the Adjacent Channel Leakage Ratio (ACLR) single sideband can be improved by 11 dB or double sideband is improved by 6 dB, with a total feedback delay of 6 ns. Better performance can be expected by integrating this method into the PA designing stage with a reduced feedback delay.
In this paper, a dual-channel nonlinear feedback architecture is proposed to suppress nonlinear distortion of Power Amplifiers (PA) in the analog domain to improve PA linearity and reduce adjacent channel leakage. In this architecture, a nonlinear extraction loop and a feedback adjustment loop are included to suppress the nonlinearity. First, in the nonlinear extraction loop, the PA input and output signals are extracted by a coupler and aligned with amplitude and phase. Then the two signals are combined to cancel the linear signal and obtain the nonlinear distortion generated by PA. Next, in the feedback adjustment loop, two independent analog channels are used to modify the amplitude and delay of the extracted nonlinear signal before injecting into the PA input port. The delays of these two channels can be finely tuned to ensure the whole feedback structure to behave like a second-order Delta-Sigma modulator, and it shows better distortion suppression performance compared with the single-channel nonlinear feedback architecture. By configuring the feedback channel parameters through the proposed method, flexible suppression of nonlinear distortion at different target frequency ranges can be achieved. An experimental platform using a commercial PA chip CMPA0060002F is designed for verification. For a test signal with a bandwidth of 40 MHz and a carrier frequency of 780 MHz, under the current hardware feedback delay of 6 ns, the Adjacent Channel Leakage Ratio (ACLR) single sideband can be improved by 11 dB or double sideband is improved by 6 dB, with a total feedback delay of 6 ns. Better performance can be expected by integrating this method into the PA designing stage with a reduced feedback delay.
2023, 45(12): 4218-4227.
doi: 10.11999/JEIT221269
Abstract:
The warehousing industry is striding forward in the direction of intelligence. Still, the challenges of in-storage and out-storage, caused by the lack of location information of goods indoors, hinder it. To achieve accurate localization of goods, a 3-D localization method based on wireless tags is proposed. The designed tags are mounted on the goods, reflecting the Orthogonal Frequency Division Multiplexing (OFDM) signals from the transmitter. The receiver with a Uniform Planar Array (UPA) as receiving antenna receives the signals and gets the channel estimates on multiple antenna channels. Then, the multi-dimensional wireless channel parameters are obtained using the two-step sparse recovery algorithm. An optimization problem for solving the unknown tags' locations is built according to the geometric locations of the receiver, transmitter, and tags in 3-D space. Finally, the swarm intelligence method is utilized to search accurately the tags' locations. The tag prototype, OFDM transmitter, and receiver are realized to validate the system based on the proposed scheme. Experimental results demonstrate that the system can achieve a 3-D median accuracy of 0.53 m.
The warehousing industry is striding forward in the direction of intelligence. Still, the challenges of in-storage and out-storage, caused by the lack of location information of goods indoors, hinder it. To achieve accurate localization of goods, a 3-D localization method based on wireless tags is proposed. The designed tags are mounted on the goods, reflecting the Orthogonal Frequency Division Multiplexing (OFDM) signals from the transmitter. The receiver with a Uniform Planar Array (UPA) as receiving antenna receives the signals and gets the channel estimates on multiple antenna channels. Then, the multi-dimensional wireless channel parameters are obtained using the two-step sparse recovery algorithm. An optimization problem for solving the unknown tags' locations is built according to the geometric locations of the receiver, transmitter, and tags in 3-D space. Finally, the swarm intelligence method is utilized to search accurately the tags' locations. The tag prototype, OFDM transmitter, and receiver are realized to validate the system based on the proposed scheme. Experimental results demonstrate that the system can achieve a 3-D median accuracy of 0.53 m.
2023, 45(12): 4228-4234.
doi: 10.11999/JEIT220934
Abstract:
Millimeter wave Multiple Input and Multiple Output (MIMO) channel exhibits beam sparsity and high directivity in the beam domain, and the beam domain channel pattern is highly correlated with the terminal position. In this paper, the beam domain channel pattern is regarded as a channel fingerprint. A channel fingerprint-based identity spoofing attacks detection scheme is proposed for millimeter-wave MIMO systems. The identity authentication problem is modeled as a binary classification problem of the corresponding channel fingerprint. Then, the supervised learning Support Vector Machine (SVM) algorithm is employed to solve the classification problem. In order to achieve good classification effect, different similarity indexes on channel fingerprint are compared based on the numerical analysis of the beam domain, and the one with the best classification effect is selected as the final classification feature to train the classifier model. The simulation results show that the proposed scheme has good authentication performance even under low signal-to-noise ratio conditions. Compared with the existing relative schemes, the detection accuracy is significantly improved.
Millimeter wave Multiple Input and Multiple Output (MIMO) channel exhibits beam sparsity and high directivity in the beam domain, and the beam domain channel pattern is highly correlated with the terminal position. In this paper, the beam domain channel pattern is regarded as a channel fingerprint. A channel fingerprint-based identity spoofing attacks detection scheme is proposed for millimeter-wave MIMO systems. The identity authentication problem is modeled as a binary classification problem of the corresponding channel fingerprint. Then, the supervised learning Support Vector Machine (SVM) algorithm is employed to solve the classification problem. In order to achieve good classification effect, different similarity indexes on channel fingerprint are compared based on the numerical analysis of the beam domain, and the one with the best classification effect is selected as the final classification feature to train the classifier model. The simulation results show that the proposed scheme has good authentication performance even under low signal-to-noise ratio conditions. Compared with the existing relative schemes, the detection accuracy is significantly improved.
2023, 45(12): 4235-4244.
doi: 10.11999/JEIT221193
Abstract:
Federated learning allows the effective flow of data value without leaving the local data, which is considered to be an effective way to balance data sharing and privacy protection in the Internet of Things (IoT) scenario. However, federated learning systems are vulnerable to Byzantine attacks and inference attacks, leading to the robustness of the system and the privacy of the data being compromised. The data heterogeneity and resource bottlenecks of IoT devices also pose significant challenges to the design of privacy-preserving and Byzantine-robust algorithms. In this paper, a data resampling of robust aggregation method Re-Sim applicable to heterogeneous IoT is proposed, which achieves robust aggregation by measuring directional similarity and normalized update magnitude, and uses data resampling to enhance robustness in data heterogeneous environments. Meanwhile, a Lightweight Security Aggregation (LSA) protocol is proposed to ensure data privacy while taking into account model robustness, accuracy and computational overhead, and the privacy of the protocol is theoretically analyzed. Simulation results show that the proposed scheme can effectively resist Byzantine attacks and inference attacks in the case of data heterogeneity. The proposed scheme improves the accuracy by 1%～3% compared to the baseline method, while reducing the client-side computational overhead by 79%.
Federated learning allows the effective flow of data value without leaving the local data, which is considered to be an effective way to balance data sharing and privacy protection in the Internet of Things (IoT) scenario. However, federated learning systems are vulnerable to Byzantine attacks and inference attacks, leading to the robustness of the system and the privacy of the data being compromised. The data heterogeneity and resource bottlenecks of IoT devices also pose significant challenges to the design of privacy-preserving and Byzantine-robust algorithms. In this paper, a data resampling of robust aggregation method Re-Sim applicable to heterogeneous IoT is proposed, which achieves robust aggregation by measuring directional similarity and normalized update magnitude, and uses data resampling to enhance robustness in data heterogeneous environments. Meanwhile, a Lightweight Security Aggregation (LSA) protocol is proposed to ensure data privacy while taking into account model robustness, accuracy and computational overhead, and the privacy of the protocol is theoretically analyzed. Simulation results show that the proposed scheme can effectively resist Byzantine attacks and inference attacks in the case of data heterogeneity. The proposed scheme improves the accuracy by 1%～3% compared to the baseline method, while reducing the client-side computational overhead by 79%.
2023, 45(12): 4245-4253.
doi: 10.11999/JEIT221386
Abstract:
Considering how to solve the problem of insufficient driving timeliness under the constraint of limited on-board energy resources of new energy vehicles, a distributed energy trading mechanism under the blockchain is proposed. Firstly, a new energy vehicle energy trading network based on the blockchain is built, and the privacy of energy trading through the Proof of Reputation (PoR) consensus mechanism is ensured. Then, based on the convergent spider web, a nonlinear pricing negotiation algorithm is designed, which combines with the blockchain technology to store the vehicle reputation database in a distributed way, to ensure that both energy trading parties can at least obtain the optimal pricing under the condition of meeting the weak Pareto effect. Finally, through simulation, the effectiveness and convergence of the proposed algorithm under the blockchain are verified, and the optimal step size and its coefficients of the algorithm are obtained.
Considering how to solve the problem of insufficient driving timeliness under the constraint of limited on-board energy resources of new energy vehicles, a distributed energy trading mechanism under the blockchain is proposed. Firstly, a new energy vehicle energy trading network based on the blockchain is built, and the privacy of energy trading through the Proof of Reputation (PoR) consensus mechanism is ensured. Then, based on the convergent spider web, a nonlinear pricing negotiation algorithm is designed, which combines with the blockchain technology to store the vehicle reputation database in a distributed way, to ensure that both energy trading parties can at least obtain the optimal pricing under the condition of meeting the weak Pareto effect. Finally, through simulation, the effectiveness and convergence of the proposed algorithm under the blockchain are verified, and the optimal step size and its coefficients of the algorithm are obtained.
2023, 45(12): 4254-4261.
doi: 10.11999/JEIT221377
Abstract:
Future cognitive Internet of Things (IoT) applications will rely on time-sensitive short-packet information to control and monitor. This paper analyzes the information freshness of secondary users in the short-packet two-way relay system. In this paper, the age of information (AoI) is used as a performance metric to measure the freshness of information. Based on the short-packet communications theory, the expressions of the packet error rate and the average peak AoI (PAoI) are derived. In addition, the performance under the high signal-to-noise ratio regime is analyzed. Due to imperfect sensing, an alternate iterative optimization algorithm is used to jointly optimize the spectrum sensing time and short-packet length to minimize the average weighted sum PAoI. The simulation results verify the correctness of the theoretical analysis. And it shows that there is a trade-off relationship between the spectrum sensing time and short-packet length, and the proposed algorithm can minimize the average weighted sum PAoI.
Future cognitive Internet of Things (IoT) applications will rely on time-sensitive short-packet information to control and monitor. This paper analyzes the information freshness of secondary users in the short-packet two-way relay system. In this paper, the age of information (AoI) is used as a performance metric to measure the freshness of information. Based on the short-packet communications theory, the expressions of the packet error rate and the average peak AoI (PAoI) are derived. In addition, the performance under the high signal-to-noise ratio regime is analyzed. Due to imperfect sensing, an alternate iterative optimization algorithm is used to jointly optimize the spectrum sensing time and short-packet length to minimize the average weighted sum PAoI. The simulation results verify the correctness of the theoretical analysis. And it shows that there is a trade-off relationship between the spectrum sensing time and short-packet length, and the proposed algorithm can minimize the average weighted sum PAoI.
2023, 45(12): 4262-4270.
doi: 10.11999/JEIT221468
Abstract:
In an extra large scale Multiple-Input Multiple-Output(MIMO) system where the Visibility Regions(VR) of different users are overlapping, the ergodic sum-rate is maximized by designing power allocation. Specifically, one base station equipped with an extra large scale array serves multiple users equipped with single-antenna, and their VRs are overlapped with adjacent users’. To reduce the inter-users interference and precoding complexity, the base station array is divided into several subarrays by the VR distributions, and then the regularized zero forcing precoding is employed for different subarray respectively. Furthermore, by exploiting the statistical channel state information, an approximation of the ergodic sum-rate is derived based on the large-dimensional random matrix theory. Based on the approximations, an optimal power allocation solution for different users is given in closed-form. Simulations illustrate that the proposed approximation fits the ergodic results well, and the proposed power allocation method can effectively improve system performances.
In an extra large scale Multiple-Input Multiple-Output(MIMO) system where the Visibility Regions(VR) of different users are overlapping, the ergodic sum-rate is maximized by designing power allocation. Specifically, one base station equipped with an extra large scale array serves multiple users equipped with single-antenna, and their VRs are overlapped with adjacent users’. To reduce the inter-users interference and precoding complexity, the base station array is divided into several subarrays by the VR distributions, and then the regularized zero forcing precoding is employed for different subarray respectively. Furthermore, by exploiting the statistical channel state information, an approximation of the ergodic sum-rate is derived based on the large-dimensional random matrix theory. Based on the approximations, an optimal power allocation solution for different users is given in closed-form. Simulations illustrate that the proposed approximation fits the ergodic results well, and the proposed power allocation method can effectively improve system performances.
Research on Wired and Wireless Time Slots Converged Scheduling Scheme for Satellite Formation Flying
2023, 45(12): 4271-4279.
doi: 10.11999/JEIT220916
Abstract:
Considering the uncertainty of the forwarding delay of time sensitive missions on the satellite caused by the difference of the transmission rate and scheduling mechanism between the intra-satellite wired and inter-satellite wireless link in the satellite formation flying scenario, a wired and wireless converged time slot scheduling scheme is proposed. Firstly, the inter-satellite wireless link transmission rate, inter-satellite wireless scheduling and intra-satellite wired scheduling are constructed respectively. Secondly, the forwarding delay analysis model of the wired and wireless converged scheduling on the satellite is established by considering the transmission rate and the time slot position relationship between the wired and wireless link. Finally, to ensure the stability of delay when the time sensitive traffic is transmitted on the satellite each time, the converged scheduling optimization goal with the minimum jitter is constructed based on the delay analysis model, and the genetic tabu search algorithm is introduced to solve the problem. Simulation results indicate that, compared with the non-converged scheduling scheme, the jitter of the proposed converged scheduling scheme is not higher than 40 μs, and the forwarding delay is reduced by an average of 20%.
Considering the uncertainty of the forwarding delay of time sensitive missions on the satellite caused by the difference of the transmission rate and scheduling mechanism between the intra-satellite wired and inter-satellite wireless link in the satellite formation flying scenario, a wired and wireless converged time slot scheduling scheme is proposed. Firstly, the inter-satellite wireless link transmission rate, inter-satellite wireless scheduling and intra-satellite wired scheduling are constructed respectively. Secondly, the forwarding delay analysis model of the wired and wireless converged scheduling on the satellite is established by considering the transmission rate and the time slot position relationship between the wired and wireless link. Finally, to ensure the stability of delay when the time sensitive traffic is transmitted on the satellite each time, the converged scheduling optimization goal with the minimum jitter is constructed based on the delay analysis model, and the genetic tabu search algorithm is introduced to solve the problem. Simulation results indicate that, compared with the non-converged scheduling scheme, the jitter of the proposed converged scheduling scheme is not higher than 40 μs, and the forwarding delay is reduced by an average of 20%.
2023, 45(12): 4280-4291.
doi: 10.11999/JEIT221247
Abstract:
With the widespread adoption of 5G technology, network hyper-density deployment has become an inevitable trend. While achieving high traffic density and high peak rate performance, ultra-dense heterogeneous wireless networks pose challenges to traditional network switching algorithms. Terminals moving at variable speeds will face more frequent switching problems, which will lead to a much higher frequency of ping-pong effects, thus affecting the user experience of the network. To address these issues, a network switching algorithm based on terminal trajectory prediction is proposed, which is applicable to both vertical and horizontal switching for all types of users in high-density wireless networks. Firstly, to predict the mobile trajectory more accurately, a prediction method based on fuzzy kernel clustering and Long Short-Term Memory (LSTM) neural networks is proposed, which can effectively predict the short-term mobile trajectory of user terminals under different mobile modes. Afterwards, two sets of candidate networks are obtained based on the current and predicted positions of the user, and the network switching timing is judged by the candidate set swapping algorithm and indicator threshold; When the switching is triggered, the emperor penguin algorithm is used to select optimally the network at the time of switching. The simulation results show that the trajectory prediction algorithm proposed has higher accuracy compared to other types of time series prediction algorithms. At the same time, compared with the comparison algorithms, the proposed network switching algorithm has a moderate number of switches, which avoids effectively the ping-pong effect and improves the network quality of user connections.
With the widespread adoption of 5G technology, network hyper-density deployment has become an inevitable trend. While achieving high traffic density and high peak rate performance, ultra-dense heterogeneous wireless networks pose challenges to traditional network switching algorithms. Terminals moving at variable speeds will face more frequent switching problems, which will lead to a much higher frequency of ping-pong effects, thus affecting the user experience of the network. To address these issues, a network switching algorithm based on terminal trajectory prediction is proposed, which is applicable to both vertical and horizontal switching for all types of users in high-density wireless networks. Firstly, to predict the mobile trajectory more accurately, a prediction method based on fuzzy kernel clustering and Long Short-Term Memory (LSTM) neural networks is proposed, which can effectively predict the short-term mobile trajectory of user terminals under different mobile modes. Afterwards, two sets of candidate networks are obtained based on the current and predicted positions of the user, and the network switching timing is judged by the candidate set swapping algorithm and indicator threshold; When the switching is triggered, the emperor penguin algorithm is used to select optimally the network at the time of switching. The simulation results show that the trajectory prediction algorithm proposed has higher accuracy compared to other types of time series prediction algorithms. At the same time, compared with the comparison algorithms, the proposed network switching algorithm has a moderate number of switches, which avoids effectively the ping-pong effect and improves the network quality of user connections.
2023, 45(12): 4292-4300.
doi: 10.11999/JEIT221297
Abstract:
Focusing on the difficulty of jointly optimizing lighting and throughput in indoor cell-free Visible Light Communication (VLC) network, a method of VLC Cooperative User Access and Dynamic Power Adjustment (VCUA-DPA) is proposed. In the user access stage, by considering the network load balance, user rate demands and the limitation of the number of cooperative cells, a user access algorithm through user and VLC cooperating is designed in this paper. In the power allocation stage, a dynamic power adjustment algorithm is designed to optimize jointly illumination uniformity and system throughput. Simulation results show that the proposed VCUA-DPA can significantly improve the system throughput and optimize the indoor illumination uniformity.
Focusing on the difficulty of jointly optimizing lighting and throughput in indoor cell-free Visible Light Communication (VLC) network, a method of VLC Cooperative User Access and Dynamic Power Adjustment (VCUA-DPA) is proposed. In the user access stage, by considering the network load balance, user rate demands and the limitation of the number of cooperative cells, a user access algorithm through user and VLC cooperating is designed in this paper. In the power allocation stage, a dynamic power adjustment algorithm is designed to optimize jointly illumination uniformity and system throughput. Simulation results show that the proposed VCUA-DPA can significantly improve the system throughput and optimize the indoor illumination uniformity.
2023, 45(12): 4301-4307.
doi: 10.11999/JEIT221281
Abstract:
A resource allocation scheme to minimize the total power is proposed for IRS-Assisted single cluster uplink Non-Orthogonal Multiple Access (NOMA) systems. Firstly, the optimization problem is constructed to minimize the total power, and the parameters are power for each user and Intelligent Reflecting Surface (IRS) phase shifts. Secondly, the relationship among the power required by single user, the channel, the user's rate requirements and IRS phase shifts is deduced. The joint optimization problem of power and phase shifts is decomposed into several sub-optimization problems with a single parameter. Then, all IRS phase shifts are solved by an iterative method, and the sub-optimization problems with a single parameter are solved in turn during each iteration. Finally, the minimum power required by each user is calculated based on the IRS phase shifts obtained by iteration. Simulation results show that the total power of the proposed scheme is lower than that of existing schemes in the same scenario.
A resource allocation scheme to minimize the total power is proposed for IRS-Assisted single cluster uplink Non-Orthogonal Multiple Access (NOMA) systems. Firstly, the optimization problem is constructed to minimize the total power, and the parameters are power for each user and Intelligent Reflecting Surface (IRS) phase shifts. Secondly, the relationship among the power required by single user, the channel, the user's rate requirements and IRS phase shifts is deduced. The joint optimization problem of power and phase shifts is decomposed into several sub-optimization problems with a single parameter. Then, all IRS phase shifts are solved by an iterative method, and the sub-optimization problems with a single parameter are solved in turn during each iteration. Finally, the minimum power required by each user is calculated based on the IRS phase shifts obtained by iteration. Simulation results show that the total power of the proposed scheme is lower than that of existing schemes in the same scenario.
2023, 45(12): 4308-4316.
doi: 10.11999/JEIT221046
Abstract:
The temporal and spatial constraints of spatio-temporal coverage tasks make it difficult to utilize the traditional single-participant model. Therefore, a task allocation method based on group collaboration in mobile crowdsensing is proposed to replace the traditional single participant mode with group mode. A task allocation framework for hierarchical group collaboration and a preference-aware social group generation method is proposed, In addition, social groups are generated by introducing social relationships to improve the task completion rate. A task-group matching method for utility optimization is proposed, and the network flow theory is used to perform group-task matching to ensure the maximum utility of the platform. Simulation results show that the proposed method can improve the task completion rate and platform utility.
The temporal and spatial constraints of spatio-temporal coverage tasks make it difficult to utilize the traditional single-participant model. Therefore, a task allocation method based on group collaboration in mobile crowdsensing is proposed to replace the traditional single participant mode with group mode. A task allocation framework for hierarchical group collaboration and a preference-aware social group generation method is proposed, In addition, social groups are generated by introducing social relationships to improve the task completion rate. A task-group matching method for utility optimization is proposed, and the network flow theory is used to perform group-task matching to ensure the maximum utility of the platform. Simulation results show that the proposed method can improve the task completion rate and platform utility.
2023, 45(12): 4317-4326.
doi: 10.11999/JEIT221084
Abstract:
This paper studies the detection and suppression mechanisms of fly synchronization of Unmanned Aerial Vehicle (UAV) swarm. Fly synchronization process is viewed as emergence in the complex system. A detection algorithm is proposed based on emergence identification with double thresholds. By simultaneously monitoring the entropy difference of flight synchronization process and network connectivity of the target system, the misjudgment of existing algorithms caused by ignoring the network status is overcomed, and the occurrence, achievement, or failure of fly synchronization is accurately identified, which provides a solid prerequisite for the timing control of the suppression mechanism. In-band radio interference behavior is designed under the constraint of average power. The interference behavior modeled from the perspective of degrading the target system’s communication capacity and the effect is analyzed through simulations. It is found that low-intensity continuous interference can effectively delay the fly synchronization process and prolong the time of that. What’s more, it has better concealment. Medium-intensity continuous interference can rapidly stop that process. Based on the above perception, for the first time, countermeasures for the UAV swarm’s fly synchronization are designed according to different operational intentions of delay and disruption. Simulation results show the effectiveness of the countermeasures.
This paper studies the detection and suppression mechanisms of fly synchronization of Unmanned Aerial Vehicle (UAV) swarm. Fly synchronization process is viewed as emergence in the complex system. A detection algorithm is proposed based on emergence identification with double thresholds. By simultaneously monitoring the entropy difference of flight synchronization process and network connectivity of the target system, the misjudgment of existing algorithms caused by ignoring the network status is overcomed, and the occurrence, achievement, or failure of fly synchronization is accurately identified, which provides a solid prerequisite for the timing control of the suppression mechanism. In-band radio interference behavior is designed under the constraint of average power. The interference behavior modeled from the perspective of degrading the target system’s communication capacity and the effect is analyzed through simulations. It is found that low-intensity continuous interference can effectively delay the fly synchronization process and prolong the time of that. What’s more, it has better concealment. Medium-intensity continuous interference can rapidly stop that process. Based on the above perception, for the first time, countermeasures for the UAV swarm’s fly synchronization are designed according to different operational intentions of delay and disruption. Simulation results show the effectiveness of the countermeasures.
2023, 45(12): 4327-4335.
doi: 10.11999/JEIT221381
Abstract:
Network virtualization is a significant technology for future network development. A method for Satellite Virtual Network (SVN) admission control is proposed in this paper to address the problem that user Quality of Service (QoS) may be severely affected in SVN, which can effectively guarantee user QoS by limiting the number of SVNs embedded in the satellite physical network. Specifically, firstly, a two-stage SVN embedding mechanism is proposed, which decouples short-term resource allocation from long-term admission control and resource leasing. Secondly, considering the uncertainty of traffic demand due to time-varying user arrival rate and the uncertainty of system capacity due to the highly dynamic nature of satellite network topology, the admission control and resource leasing problems in the first stage are described as robust optimization problems, which are then transformed into convex problems using the Bernstein approximation for solution. Finally, the article transforms the resource allocation problem in the second stage into a convex problem that maximizes fair bandwidth allocation for solution. The simulation results demonstrate the effectiveness of the proposed scheme in this paper.
Network virtualization is a significant technology for future network development. A method for Satellite Virtual Network (SVN) admission control is proposed in this paper to address the problem that user Quality of Service (QoS) may be severely affected in SVN, which can effectively guarantee user QoS by limiting the number of SVNs embedded in the satellite physical network. Specifically, firstly, a two-stage SVN embedding mechanism is proposed, which decouples short-term resource allocation from long-term admission control and resource leasing. Secondly, considering the uncertainty of traffic demand due to time-varying user arrival rate and the uncertainty of system capacity due to the highly dynamic nature of satellite network topology, the admission control and resource leasing problems in the first stage are described as robust optimization problems, which are then transformed into convex problems using the Bernstein approximation for solution. Finally, the article transforms the resource allocation problem in the second stage into a convex problem that maximizes fair bandwidth allocation for solution. The simulation results demonstrate the effectiveness of the proposed scheme in this paper.
2023, 45(12): 4336-4344.
doi: 10.11999/JEIT221163
Abstract:
During the observation of micro-motion targets in space, high resolution radar collects the narrowband and wideband echoes simultaneously. This paper proposes a fusion method based on a Sparse Auto-Encoder (SAE) for recognizing space micro-motion targets to exploit their rich electromagnetic scattering, shape, structure, and motion information. In the training phase, the proposed method extracts the hierarchical features from High Resolution Range Profiles (HRRP), Joint Time-Frequency (JTF) images, and Range-Instantaneous-Doppler (RID) images using Convolution Neural Networks (CNN). The joint feature vector is then created by concatenating the relevant deep features, and SAE learns autonomously its hidden features unsupervised. After that, the decoder is removed and the Softmax classifier is introduced after the encoder to create the recognition network. Finally, parameters of the optimized SAE network are used for the initialization of the recognition network, which is then fine-tuned by the joint feature vectors of training samples. In the test phase, the trained recognition network is supplied directly with the joint feature vectors of the test samples recovered by CNN to produce the fusion recognition results. Experimental results of simulated EM data under different conditions show the efficacy and robustness of the proposed method.
During the observation of micro-motion targets in space, high resolution radar collects the narrowband and wideband echoes simultaneously. This paper proposes a fusion method based on a Sparse Auto-Encoder (SAE) for recognizing space micro-motion targets to exploit their rich electromagnetic scattering, shape, structure, and motion information. In the training phase, the proposed method extracts the hierarchical features from High Resolution Range Profiles (HRRP), Joint Time-Frequency (JTF) images, and Range-Instantaneous-Doppler (RID) images using Convolution Neural Networks (CNN). The joint feature vector is then created by concatenating the relevant deep features, and SAE learns autonomously its hidden features unsupervised. After that, the decoder is removed and the Softmax classifier is introduced after the encoder to create the recognition network. Finally, parameters of the optimized SAE network are used for the initialization of the recognition network, which is then fine-tuned by the joint feature vectors of training samples. In the test phase, the trained recognition network is supplied directly with the joint feature vectors of the test samples recovered by CNN to produce the fusion recognition results. Experimental results of simulated EM data under different conditions show the efficacy and robustness of the proposed method.
2023, 45(12): 4345-4351.
doi: 10.11999/JEIT221320
Abstract:
Restricted to the spatial layout of complex systems, the cables used in these systems are usually harness structures, and have bent and stereoscopic configurations. At present, efficient time-domain modeling and analysis methods for the coupling of harness cable with Bent and Stereoscopic Configurations (BSCs) are still rare. Therefore, an efficient time-domain hybrid method, consisting of the Finite Difference Time Domain (FDTD) method, Transmission Line (TL) equations, adaptive cable mesh technique, interpolation techniques and charge conservation law, is studied to achieve the fast and synchronous calculations of space electromagnetic field radiation and the coupling responses of harness cable with BSCs. Firstly, the structure of the harness cable with BSCs is decomposed into multiple independent sub harness cable according to the bending nodes. Then, the coupling model of each sub harness cable is constructed by the TL equations, in which the adaptive cable mesh technique and some interpolation techniques are employed to compute the distribution sources of the TL equations, and the FDTD is applied to solve the transient responses on the sub harness cable. Finally, the equivalent circuit model of the bending nodes are constructed by the charge conservation law, and the voltages at the nodes are solved and fed back to these sub harness cables to realize the interference signal transmission between these cables. To verify the accuracy and efficiency of the proposed method, two coupling problems of the harness cable with BSCs in the environments of free space and shielding enclosure are solved by this method, CST and Finite Difference Time Domain-Simulation Program with Integrated Circuit Emphasis (FDTD-SPICE) method, which are compared in the calculation precision and time consumption.
Restricted to the spatial layout of complex systems, the cables used in these systems are usually harness structures, and have bent and stereoscopic configurations. At present, efficient time-domain modeling and analysis methods for the coupling of harness cable with Bent and Stereoscopic Configurations (BSCs) are still rare. Therefore, an efficient time-domain hybrid method, consisting of the Finite Difference Time Domain (FDTD) method, Transmission Line (TL) equations, adaptive cable mesh technique, interpolation techniques and charge conservation law, is studied to achieve the fast and synchronous calculations of space electromagnetic field radiation and the coupling responses of harness cable with BSCs. Firstly, the structure of the harness cable with BSCs is decomposed into multiple independent sub harness cable according to the bending nodes. Then, the coupling model of each sub harness cable is constructed by the TL equations, in which the adaptive cable mesh technique and some interpolation techniques are employed to compute the distribution sources of the TL equations, and the FDTD is applied to solve the transient responses on the sub harness cable. Finally, the equivalent circuit model of the bending nodes are constructed by the charge conservation law, and the voltages at the nodes are solved and fed back to these sub harness cables to realize the interference signal transmission between these cables. To verify the accuracy and efficiency of the proposed method, two coupling problems of the harness cable with BSCs in the environments of free space and shielding enclosure are solved by this method, CST and Finite Difference Time Domain-Simulation Program with Integrated Circuit Emphasis (FDTD-SPICE) method, which are compared in the calculation precision and time consumption.
2023, 45(12): 4352-4360.
doi: 10.11999/JEIT221385
Abstract:
The single frequency or modulated frequency waveform employed in active Sound Navigation and Ranging (SONAR) systems yields inferior range or velocity resolution. This deficiency impairs signal detection and estimation in reverberation background. Previous work demonstrates that “V” and “W” type Hyperbolic Frequency Modulation (HFM) combination waveforms substantially enhance range–velocity resolution while simultaneously reducing reverberation. The W-HFM waveform serves as an effective solution to the false-target problem inherent in the V-HFM waveform when applied to multiple-target scenarios. However, the complex design and extensive computation required pose considerable challenges. Thus, the ridge slope equation of HFM combination waveforms is derived to optimize their design and application. Furthermore, the minimum no false alarm distance index of the V-HFM waveform is proposed, and its applicability in multiple-target scenarios is analyzed. Additionally, an optimized waveform design scheme is proposed, using the typical W-HFM waveform as an example, which can serve as a guide for engineering applications. Tank experimental data reveal that the HFM combination waveform achieves high range–velocity resolution, reverberation is reduced by more than 5 dB, and the W-HFM waveform suppresses false target interference.
The single frequency or modulated frequency waveform employed in active Sound Navigation and Ranging (SONAR) systems yields inferior range or velocity resolution. This deficiency impairs signal detection and estimation in reverberation background. Previous work demonstrates that “V” and “W” type Hyperbolic Frequency Modulation (HFM) combination waveforms substantially enhance range–velocity resolution while simultaneously reducing reverberation. The W-HFM waveform serves as an effective solution to the false-target problem inherent in the V-HFM waveform when applied to multiple-target scenarios. However, the complex design and extensive computation required pose considerable challenges. Thus, the ridge slope equation of HFM combination waveforms is derived to optimize their design and application. Furthermore, the minimum no false alarm distance index of the V-HFM waveform is proposed, and its applicability in multiple-target scenarios is analyzed. Additionally, an optimized waveform design scheme is proposed, using the typical W-HFM waveform as an example, which can serve as a guide for engineering applications. Tank experimental data reveal that the HFM combination waveform achieves high range–velocity resolution, reverberation is reduced by more than 5 dB, and the W-HFM waveform suppresses false target interference.
2023, 45(12): 4361-4370.
doi: 10.11999/JEIT221013
Abstract:
Radar-communication integration can realize both target detection and wireless communication, so as to reduce electromagnetic interference and improve spectrum utilization. The integrated signal design of radar and communication is the key to the realization of integrated technology. The common radar communication integration signals based on Orthogonal Frequency Division Multiplexing (OFDM) have the problems of frequency offset sensitivity and excessive out-of-band radiation, so they are not suitable for high-dynamic applications. Considering the advantages of Filter Bank MultiCarrier (FBMC) signal with high Doppler tolerance and low out-of-band leakage, the time-frequency locations of radar and communication subcarriers are optimized under the framework of FBMC, and a design method of FBMC comb spectrum radar-communication integrated signal is proposed. Since there is inherent interference between FBMC signal carriers and symbols, which leads to inaccurate channel estimation and is not suitable for fast time-varying channels, an interleaved comb-shaped auxiliary pilot structure is designed to eliminate inherent interference and achieve channel tracking. In addition, the radar complex signal in the integrated signal introduces real interference to the communication signal. A real interference compensation algorithm is proposed based on interference utilization, which is used to restore the communication signal. The simulation results show that under the fast time-varying channel, the designed radar-communication integrated signal has a lower bit error rate and better radar detection performance in the process of high data rate transmission.
Radar-communication integration can realize both target detection and wireless communication, so as to reduce electromagnetic interference and improve spectrum utilization. The integrated signal design of radar and communication is the key to the realization of integrated technology. The common radar communication integration signals based on Orthogonal Frequency Division Multiplexing (OFDM) have the problems of frequency offset sensitivity and excessive out-of-band radiation, so they are not suitable for high-dynamic applications. Considering the advantages of Filter Bank MultiCarrier (FBMC) signal with high Doppler tolerance and low out-of-band leakage, the time-frequency locations of radar and communication subcarriers are optimized under the framework of FBMC, and a design method of FBMC comb spectrum radar-communication integrated signal is proposed. Since there is inherent interference between FBMC signal carriers and symbols, which leads to inaccurate channel estimation and is not suitable for fast time-varying channels, an interleaved comb-shaped auxiliary pilot structure is designed to eliminate inherent interference and achieve channel tracking. In addition, the radar complex signal in the integrated signal introduces real interference to the communication signal. A real interference compensation algorithm is proposed based on interference utilization, which is used to restore the communication signal. The simulation results show that under the fast time-varying channel, the designed radar-communication integrated signal has a lower bit error rate and better radar detection performance in the process of high data rate transmission.
2023, 45(12): 4371-4381.
doi: 10.11999/JEIT221400
Abstract:
Text-to-image generation is a comprehensive task that combines the fields of Computer Vision (CV) and Natural Language Processing (NLP). Research on the methods of text to image based on Generative Adversarial Networks (GANs) continues to grow in popularity and have made some progress, but the methods of GANs model suffer from training instability. To address this problem, a text-to-image generation model based on Diffusion Wasserstein Generative Adversarial Networks (D-WGAN) is proposed, which generates high quality and diverse images and enables stable training process by feeding randomly sampled instance noise from the diffusion process into the discriminator. Considering the high cost of sampling the diffusion process, a stochastic differentiation method is introduced to simplify the sampling process. In order to align further the information of text and image, Contrastive Language-Image Pre-training (CLIP) model is introduced to obtain the cross-modal mapping relationship between text and image information, so as to improve the consistency of text and image. Experimental results on the MSCOCO and CUB-200 datasets show that D-WGAN achieves stable training while reducing Fréchet Inception Distance (FID) scores by 16.43% and 1.97%, respectively, and improving Inception Score (IS) scores by 3.38% and 30.95%, respectively. These results indicate that D-WGAN can generate higher quality images and has more practical value.
Text-to-image generation is a comprehensive task that combines the fields of Computer Vision (CV) and Natural Language Processing (NLP). Research on the methods of text to image based on Generative Adversarial Networks (GANs) continues to grow in popularity and have made some progress, but the methods of GANs model suffer from training instability. To address this problem, a text-to-image generation model based on Diffusion Wasserstein Generative Adversarial Networks (D-WGAN) is proposed, which generates high quality and diverse images and enables stable training process by feeding randomly sampled instance noise from the diffusion process into the discriminator. Considering the high cost of sampling the diffusion process, a stochastic differentiation method is introduced to simplify the sampling process. In order to align further the information of text and image, Contrastive Language-Image Pre-training (CLIP) model is introduced to obtain the cross-modal mapping relationship between text and image information, so as to improve the consistency of text and image. Experimental results on the MSCOCO and CUB-200 datasets show that D-WGAN achieves stable training while reducing Fréchet Inception Distance (FID) scores by 16.43% and 1.97%, respectively, and improving Inception Score (IS) scores by 3.38% and 30.95%, respectively. These results indicate that D-WGAN can generate higher quality images and has more practical value.
2023, 45(12): 4382-4390.
doi: 10.11999/JEIT230553
Abstract:
To address the issue of high input Signal-to-Noise Ratio (SNR) in the spatial spectrum synthesis technique based on noise subspace projection, an improved noise subspace projection method based on spatial transformation preprocessing is proposed. First, the receiver array is uniformly split into sub-arrays to form three-dimensional space-space-frequency data. Then, three-dimensional data is projected into two-dimensional space-frequency data by spatial transformation, realizing coherent accumulation in the sub-array dimension and enhancing the input SNR after spatial transformation. Finally, the spatial spectrum synthesis is achieved by processing the two-dimensional transformed data, based on the noise subspace projection method. Numerical simulation and data processing results demonstrate that, compared with the noise subspace projection method, the proposed method decreases effectively the minimum input SNR by 6 dB while maintaining the bearing resolution, enhancing effectively the weak target detection performance of the noise subspace projection method.
To address the issue of high input Signal-to-Noise Ratio (SNR) in the spatial spectrum synthesis technique based on noise subspace projection, an improved noise subspace projection method based on spatial transformation preprocessing is proposed. First, the receiver array is uniformly split into sub-arrays to form three-dimensional space-space-frequency data. Then, three-dimensional data is projected into two-dimensional space-frequency data by spatial transformation, realizing coherent accumulation in the sub-array dimension and enhancing the input SNR after spatial transformation. Finally, the spatial spectrum synthesis is achieved by processing the two-dimensional transformed data, based on the noise subspace projection method. Numerical simulation and data processing results demonstrate that, compared with the noise subspace projection method, the proposed method decreases effectively the minimum input SNR by 6 dB while maintaining the bearing resolution, enhancing effectively the weak target detection performance of the noise subspace projection method.
2023, 45(12): 4391-4398.
doi: 10.11999/JEIT221292
Abstract:
Eight-Sided Fortress(ESF), an improved lightweight block cipher based on LBlock, has excellent software and hardware implementation efficiency. For the security of ESF, with the help of automated search tools, the algorithm is evaluated for security using the impossible differential cryptanalysis. Firstly, an impossible differential search model based on Mixed Integer Linear Programming (MILP) is built by combining the structure of ESF algorithm and the differential propagation of\begin{document}$ S $\end{document} \begin{document}$ S $\end{document} \begin{document}$ {2^{60.16}} $\end{document} \begin{document}$ {2^{67.44}} $\end{document}
Eight-Sided Fortress(ESF), an improved lightweight block cipher based on LBlock, has excellent software and hardware implementation efficiency. For the security of ESF, with the help of automated search tools, the algorithm is evaluated for security using the impossible differential cryptanalysis. Firstly, an impossible differential search model based on Mixed Integer Linear Programming (MILP) is built by combining the structure of ESF algorithm and the differential propagation of
2023, 45(12): 4399-4410.
doi: 10.11999/JEIT221251
Abstract:
Although deep neural networks can effectively improve Environmental Sound Classification (ESC) performance, they are still vulnerable to adversarial attacks. The existing adversarial defense methods are usually effective only for specific attacks and can not be adapted to different attack settings such as white-box and black-box. To improve the defense capability of ESC models in various attacking scenarios, an ESC adversarial defense method is proposed in this paper, which combines adversarial detection, adversarial training, and discriminative feature learning. This method uses an Adversarial Example Detector (AED) to detect samples input to the ESC model, and trains both the AED and ESC model simultaneously via Generative Adversarial Network (GAN), where the AED is used as the discriminator of GAN. Meanwhile, this method introduces discriminative loss functions into the adversarial training of the ESC model, so as to drive the model to learn deep features more compact within classes and more distant between classes, which helps to improve further the adversarial robustness of the model. Comparative experiments of multiple defense methods on two typical ESC datasets under white-box, adaptive white-box, and black-box attack settings are conducted. The experimental results show that by implementing a combination of multiple defense methods based on GAN, the proposed method can effectively improve the defense capability of ESC models against various attacks, and the corresponding ESC accuracy is at least 10% higher than that achieved by other defense methods. Meanwhile, it is verified that the effectiveness of the proposed method is not due to the obfuscated gradients.
Although deep neural networks can effectively improve Environmental Sound Classification (ESC) performance, they are still vulnerable to adversarial attacks. The existing adversarial defense methods are usually effective only for specific attacks and can not be adapted to different attack settings such as white-box and black-box. To improve the defense capability of ESC models in various attacking scenarios, an ESC adversarial defense method is proposed in this paper, which combines adversarial detection, adversarial training, and discriminative feature learning. This method uses an Adversarial Example Detector (AED) to detect samples input to the ESC model, and trains both the AED and ESC model simultaneously via Generative Adversarial Network (GAN), where the AED is used as the discriminator of GAN. Meanwhile, this method introduces discriminative loss functions into the adversarial training of the ESC model, so as to drive the model to learn deep features more compact within classes and more distant between classes, which helps to improve further the adversarial robustness of the model. Comparative experiments of multiple defense methods on two typical ESC datasets under white-box, adaptive white-box, and black-box attack settings are conducted. The experimental results show that by implementing a combination of multiple defense methods based on GAN, the proposed method can effectively improve the defense capability of ESC models against various attacks, and the corresponding ESC accuracy is at least 10% higher than that achieved by other defense methods. Meanwhile, it is verified that the effectiveness of the proposed method is not due to the obfuscated gradients.
2023, 45(12): 4411-4420.
doi: 10.11999/JEIT220980
Abstract:
In view of the problems that the current mainstream cross-modal image retrieval algorithm based on Convolutional Neural Network (CNN) paradigm can not extract details of ship images effectively, and the cross-modal “heterogeneous gap” is difficult to eliminate, a Discriminant Adversarial Hash Transformer (DAHT) is proposed for fast cross-modal retrieval of ship images. The network adopts dual-stream Vision Transformer(ViT) structure and relies on the self-attention mechanism of ViT to extract the discriminant features of ship images. Based on this, a Hash Token structure is designed for Hash generation. In order to eliminate the cross-modal difference of the same category image, the whole retrieval framework is trained in an adversarial way, and modal confusion is realized by modal discrimination of generated Hash codes. At the same time, a Normalized discounted cumulative gain Weighting based Discriminant Cross-modal Quintuplet Loss (NW-DCQL) is designed to maintain the semantic discrimination of different types of images. In the four types of cross-modal retrieval tasks carried out on two datasets, the proposed method achieves 9.8 %, 5.2 %, 19.7 %, and 21.6 % performance improvement compared with the suboptimal retrieval results (32 bit), and also has certain performance advantages in unimodal retrieval tasks.
In view of the problems that the current mainstream cross-modal image retrieval algorithm based on Convolutional Neural Network (CNN) paradigm can not extract details of ship images effectively, and the cross-modal “heterogeneous gap” is difficult to eliminate, a Discriminant Adversarial Hash Transformer (DAHT) is proposed for fast cross-modal retrieval of ship images. The network adopts dual-stream Vision Transformer(ViT) structure and relies on the self-attention mechanism of ViT to extract the discriminant features of ship images. Based on this, a Hash Token structure is designed for Hash generation. In order to eliminate the cross-modal difference of the same category image, the whole retrieval framework is trained in an adversarial way, and modal confusion is realized by modal discrimination of generated Hash codes. At the same time, a Normalized discounted cumulative gain Weighting based Discriminant Cross-modal Quintuplet Loss (NW-DCQL) is designed to maintain the semantic discrimination of different types of images. In the four types of cross-modal retrieval tasks carried out on two datasets, the proposed method achieves 9.8 %, 5.2 %, 19.7 %, and 21.6 % performance improvement compared with the suboptimal retrieval results (32 bit), and also has certain performance advantages in unimodal retrieval tasks.
2023, 45(12): 4421-4430.
doi: 10.11999/JEIT221300
Abstract:
Different operational states of urban rail transit usually correspond to different Intrinsic Mode Functions (IMFs) and time-scale characteristics in passenger flow time series. A combined deep learning prediction model for short-term passenger flow time series of subway is proposed based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Bidirectional Long Short Term Memory network (BiLSTM), including: mode decomposition of passenger flow time series based on the CEEMDAN algorithm. The sample entropy and hierarchical clustering are used respectively to analyze the complexity and similarity of IMFs. The IMFs are then classified, merged and reconstructed on this basis. The hyper-parameters of the model are optimized using the Tree-structured Parzen Estimator (TPE) in the Optuna framework, and the combined prediction model CEEMDAN-TPE-BiLSTM is established. Actual data are used to validate the model. The results show that the accuracy and validity indicators of the model all reach the optimum for passenger flow time series data with specific characteristics.
Different operational states of urban rail transit usually correspond to different Intrinsic Mode Functions (IMFs) and time-scale characteristics in passenger flow time series. A combined deep learning prediction model for short-term passenger flow time series of subway is proposed based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Bidirectional Long Short Term Memory network (BiLSTM), including: mode decomposition of passenger flow time series based on the CEEMDAN algorithm. The sample entropy and hierarchical clustering are used respectively to analyze the complexity and similarity of IMFs. The IMFs are then classified, merged and reconstructed on this basis. The hyper-parameters of the model are optimized using the Tree-structured Parzen Estimator (TPE) in the Optuna framework, and the combined prediction model CEEMDAN-TPE-BiLSTM is established. Actual data are used to validate the model. The results show that the accuracy and validity indicators of the model all reach the optimum for passenger flow time series data with specific characteristics.
2023, 45(12): 4431-4439.
doi: 10.11999/JEIT221097
Abstract:
A 2-dimension compressed sensing algorithm based on adaptive blocking and joint optimization Smooth l0 (SL0) norm is proposed to solve the problem of poor compression and reconstruction performance of the traditional compressed sensing model and reconstruction method. In the compression process, gray entropy and quadtree algorithm are used for adaptive blocking and sample rate allocation. At the same time, the compressed sensing model is optimized and the chaotic cyclic matrix is used as the measure matrix, which improves the compression performance. In the reconstruction process based on SL0 algorithm, a fitting function with higher steepness and a scheme combined with Quasi-Newton method and dynamic iteration are adopted to improve the reconstruction quality and efficiency. Compared with other algorithms, the peak signal to noise ratio and structural similarity index of the proposed algorithm are improved by 5.44 dB and 21.08% on average respectively. The average calculation time is only 1.59 s. Based on realizing image compression and accurate reconstruction stably and quickly, the proposed algorithm provides a new method for compressed sensing and image reconstruction.
A 2-dimension compressed sensing algorithm based on adaptive blocking and joint optimization Smooth l0 (SL0) norm is proposed to solve the problem of poor compression and reconstruction performance of the traditional compressed sensing model and reconstruction method. In the compression process, gray entropy and quadtree algorithm are used for adaptive blocking and sample rate allocation. At the same time, the compressed sensing model is optimized and the chaotic cyclic matrix is used as the measure matrix, which improves the compression performance. In the reconstruction process based on SL0 algorithm, a fitting function with higher steepness and a scheme combined with Quasi-Newton method and dynamic iteration are adopted to improve the reconstruction quality and efficiency. Compared with other algorithms, the peak signal to noise ratio and structural similarity index of the proposed algorithm are improved by 5.44 dB and 21.08% on average respectively. The average calculation time is only 1.59 s. Based on realizing image compression and accurate reconstruction stably and quickly, the proposed algorithm provides a new method for compressed sensing and image reconstruction.
2023, 45(12): 4440-4447.
doi: 10.11999/JEIT221442
Abstract:
Considering the problem that the online detection performance of the offline model will experience performance degradation when the fading dynamics of the wireless channel in the complex environment of the mine are changed, the Adaptive Detection Network (ADN) based on transfer learning is studied. The main improvement of ADN is the use of parallel networks to discretize dynamic channels to improve network generalization capabilities. The unsupervised learning method of Domain Adversarial training of Neural Network (DANN) is adopted for the online receiver signal, so as to transfer the offline training knowledge to the complex environment of the online mine and adjust the network parameters in real time to adapt to the change of channel. Finally, it realizes the adaptive signal detection in the complex environment of the mine. Experiments show that ADN obtains the diversity benefit between channels for Quadrature Phase Shift Keying (QPSK) and Quadrature Amplitude Modulation (QAM) signals in the dynamically changing Nakagami-m fading channel. The performance gradually improves with the increase of discrete channels. At high Signal-to-Noise Ratio (SNR), its performance is close to that of Convolutional Neural Network (CNN). The robustness and online detection effect of deep detection networks are significantly improved at low SNR.
Considering the problem that the online detection performance of the offline model will experience performance degradation when the fading dynamics of the wireless channel in the complex environment of the mine are changed, the Adaptive Detection Network (ADN) based on transfer learning is studied. The main improvement of ADN is the use of parallel networks to discretize dynamic channels to improve network generalization capabilities. The unsupervised learning method of Domain Adversarial training of Neural Network (DANN) is adopted for the online receiver signal, so as to transfer the offline training knowledge to the complex environment of the online mine and adjust the network parameters in real time to adapt to the change of channel. Finally, it realizes the adaptive signal detection in the complex environment of the mine. Experiments show that ADN obtains the diversity benefit between channels for Quadrature Phase Shift Keying (QPSK) and Quadrature Amplitude Modulation (QAM) signals in the dynamically changing Nakagami-m fading channel. The performance gradually improves with the increase of discrete channels. At high Signal-to-Noise Ratio (SNR), its performance is close to that of Convolutional Neural Network (CNN). The robustness and online detection effect of deep detection networks are significantly improved at low SNR.
2023, 45(12): 4448-4457.
doi: 10.11999/JEIT221205
Abstract:
Decolorization is an image compression method widely used in various fields, but few researches focus on the mutual conversion technology of color image and grayscale image. In this paper, a deep learning method is used to propose innovatively an invertible decolorization method based on variable augmentation. This method uses variable augmentation technology to ensure that the output has the same number of channels as the input variable, which satisfies the reversible characteristics of the network. Specifically, the proposed method realizes the decolorization through the forward process of the invertible neural network, and realizes the color restoration of grayscale images through the reverse process. The proposed method performs qualitative and quantitative comparisons on VOC2012, NCD, Wallpaper datasets. The experimental results show that the proposed method achieves better results in the evaluation indicators. The quality of the generated images can preserve the characteristics of brightness, color contrast and structural correlation to the greatest extent, both globally and locally.
Decolorization is an image compression method widely used in various fields, but few researches focus on the mutual conversion technology of color image and grayscale image. In this paper, a deep learning method is used to propose innovatively an invertible decolorization method based on variable augmentation. This method uses variable augmentation technology to ensure that the output has the same number of channels as the input variable, which satisfies the reversible characteristics of the network. Specifically, the proposed method realizes the decolorization through the forward process of the invertible neural network, and realizes the color restoration of grayscale images through the reverse process. The proposed method performs qualitative and quantitative comparisons on VOC2012, NCD, Wallpaper datasets. The experimental results show that the proposed method achieves better results in the evaluation indicators. The quality of the generated images can preserve the characteristics of brightness, color contrast and structural correlation to the greatest extent, both globally and locally.
2023, 45(12): 4458-4468.
doi: 10.11999/JEIT221570
Abstract:
Intelligent analysis of surveillance video data based on deep learning models can improve the cultural relics security risk prevention capabilities of cultural relics museum units. In view of the needs of cultural relics museum units to make full use of existing free and available computing resources to complete more intelligent analysis of video data, a dynamic scheduling method for video intelligent analysis tasks is proposed. The device serves as a computing node to form an edge computing power collaborative system to process intelligent video analysis tasks. In this paper, the problem to be solved is modeled as a two-dimensional multiple knapsack problem, and the method of dynamic programming is used to solve how to allocate dynamically video analysis tasks on the edge computing power collaborative system so that the security value and benefits obtained by the system execution tasks in each time period problem of maximization. The simulation results show that the proposed method can dynamically allocate video intelligent analysis tasks based on the monitoring and analysis of the current resource usage status of the system without interfering with the normal business application services of cultural relics museum units, achieving the goal of maximizing security value and benefits purpose.
Intelligent analysis of surveillance video data based on deep learning models can improve the cultural relics security risk prevention capabilities of cultural relics museum units. In view of the needs of cultural relics museum units to make full use of existing free and available computing resources to complete more intelligent analysis of video data, a dynamic scheduling method for video intelligent analysis tasks is proposed. The device serves as a computing node to form an edge computing power collaborative system to process intelligent video analysis tasks. In this paper, the problem to be solved is modeled as a two-dimensional multiple knapsack problem, and the method of dynamic programming is used to solve how to allocate dynamically video analysis tasks on the edge computing power collaborative system so that the security value and benefits obtained by the system execution tasks in each time period problem of maximization. The simulation results show that the proposed method can dynamically allocate video intelligent analysis tasks based on the monitoring and analysis of the current resource usage status of the system without interfering with the normal business application services of cultural relics museum units, achieving the goal of maximizing security value and benefits purpose.
2023, 45(12): 4469-4476.
doi: 10.11999/JEIT221549
Abstract:
A Spiking Neural Network (SNN) is a low-power neural network that simulates the dynamics of neurons in the brain, providing a feasible solution for deploying object detection tasks in high computational efficiency and low energy consumption environments. Due to the non-differentiable nature of spikes, SNN training is difficult, and a practical solution is to convert pre-trained Artificial Neural Networks (ANNs) into SNNs to improve inference ability. However, the converted SNN often suffers from performance degradation and high latency, which can not meet the high-precision localization requirements for object detection tasks. A dual error is introduced to reduce the loss of conversion performance. To simulate the ANN to SNN conversion error, the causes of errors are analyzed, and a dual error model is built. Further, the dual error model is introduced into the ANN training process so that the errors of the models before and after conversion remain consistent during training and testing, thereby reducing the loss of conversion performance. Finally, the lightweight detection algorithm YOLO is used to verify the effectiveness of the dual error model on the PASCAL VOC and MS COCO datasets.
A Spiking Neural Network (SNN) is a low-power neural network that simulates the dynamics of neurons in the brain, providing a feasible solution for deploying object detection tasks in high computational efficiency and low energy consumption environments. Due to the non-differentiable nature of spikes, SNN training is difficult, and a practical solution is to convert pre-trained Artificial Neural Networks (ANNs) into SNNs to improve inference ability. However, the converted SNN often suffers from performance degradation and high latency, which can not meet the high-precision localization requirements for object detection tasks. A dual error is introduced to reduce the loss of conversion performance. To simulate the ANN to SNN conversion error, the causes of errors are analyzed, and a dual error model is built. Further, the dual error model is introduced into the ANN training process so that the errors of the models before and after conversion remain consistent during training and testing, thereby reducing the loss of conversion performance. Finally, the lightweight detection algorithm YOLO is used to verify the effectiveness of the dual error model on the PASCAL VOC and MS COCO datasets.
Magnetic Induction Tomography of IntraCerebral Hemorrhage Based on Improved Newton-Raphson Algorithm
2023, 45(12): 4477-4488.
doi: 10.11999/JEIT221393
Abstract:
To solve the problems of over-simplified positive problem model, low image reconstruction quality, low algorithm convergence efficiency, large artifacts between lesion and background, and long time consuming in IntraCerebral Hemorrhage (ICH) Magnetic Induction Tomography (MIT), an improved Newton-Raphson (NR) algorithm for MIT of intracerebral hemorrhage is proposed. The calculation results of Linear Back Projection (LBP) algorithm are used as the iterative initial values of the improved NR algorithm, the adaptive acceleration penalty term and the L2 norm penalty term are added to the objective function to improve the efficiency of each iteration of the algorithm and reduce the artifacts of the reconstructed image. A real three-dimensional brain model including scalp, skull, cerebrospinal fluid and brain parenchyma is constructed by Comsol Multiphysics. The phase difference detection value and sensitivity matrix are simulated and calculated for subsequent image reconstruction. The proposed improved NR algorithm and five image reconstruction algorithms are used to perform magnetic induction tomography on intracerebral hemorrhage with blood loss of 24 ml, 14 ml and 2 ml at three locations, respectively. The experimental results show that the proposed algorithm has higher quality of reconstructed images than the other five algorithms. The average imaging time is only 1/3 of the NR algorithm. The higher quality image is reconstructed with fewer iterations, the image reconstruction of 2 ml intracerebral hemorrhage can be realized, which provides a new and effective algorithm for MIT detection of intracerebral hemorrhage.
To solve the problems of over-simplified positive problem model, low image reconstruction quality, low algorithm convergence efficiency, large artifacts between lesion and background, and long time consuming in IntraCerebral Hemorrhage (ICH) Magnetic Induction Tomography (MIT), an improved Newton-Raphson (NR) algorithm for MIT of intracerebral hemorrhage is proposed. The calculation results of Linear Back Projection (LBP) algorithm are used as the iterative initial values of the improved NR algorithm, the adaptive acceleration penalty term and the L2 norm penalty term are added to the objective function to improve the efficiency of each iteration of the algorithm and reduce the artifacts of the reconstructed image. A real three-dimensional brain model including scalp, skull, cerebrospinal fluid and brain parenchyma is constructed by Comsol Multiphysics. The phase difference detection value and sensitivity matrix are simulated and calculated for subsequent image reconstruction. The proposed improved NR algorithm and five image reconstruction algorithms are used to perform magnetic induction tomography on intracerebral hemorrhage with blood loss of 24 ml, 14 ml and 2 ml at three locations, respectively. The experimental results show that the proposed algorithm has higher quality of reconstructed images than the other five algorithms. The average imaging time is only 1/3 of the NR algorithm. The higher quality image is reconstructed with fewer iterations, the image reconstruction of 2 ml intracerebral hemorrhage can be realized, which provides a new and effective algorithm for MIT detection of intracerebral hemorrhage.
2023, 45(12): 4489-4498.
doi: 10.11999/JEIT221270
Abstract:
Saliency detection of light field images is a key technique in applications such as visual tracking, target detection, and image compression. However, the existing deep learning methods ignore feature differences and global contextual information when processing features, resulting in blurred saliency maps and even incomplete detection objects and difficult background suppression in scenes with similar foreground and background colors, textures, or background clutter. A context-aware cross-layer feature fusion-based saliency detection network for light field images is proposed. First, a cross-layer feature fusion module is built to select adaptively complementary components from input features to reduce feature differences and avoid inaccurate integration of features in order to more effectively fuse adjacent layer features and informative coefficients; Meanwhile, a Parallel Cascaded Feedback Decoder (PCFD) is constructed using the cross-layer feature fusion module to iteratively refine features using a multi-level feedback mechanism to avoid feature loss and dilution of high-level contextual features; Finally, a Global Context Module (GCM) generates multi-scale features to exploit the rich global context information in order to obtain the correlation between different salient regions and mitigate the dilution of high-level features. Experimental results on the latest light field dataset show that the textual method outperforms the compared methods both quantitatively and qualitatively, and is able to detect accurately complete salient objects and obtain clear saliency maps from similar front/background scenes.
Saliency detection of light field images is a key technique in applications such as visual tracking, target detection, and image compression. However, the existing deep learning methods ignore feature differences and global contextual information when processing features, resulting in blurred saliency maps and even incomplete detection objects and difficult background suppression in scenes with similar foreground and background colors, textures, or background clutter. A context-aware cross-layer feature fusion-based saliency detection network for light field images is proposed. First, a cross-layer feature fusion module is built to select adaptively complementary components from input features to reduce feature differences and avoid inaccurate integration of features in order to more effectively fuse adjacent layer features and informative coefficients; Meanwhile, a Parallel Cascaded Feedback Decoder (PCFD) is constructed using the cross-layer feature fusion module to iteratively refine features using a multi-level feedback mechanism to avoid feature loss and dilution of high-level contextual features; Finally, a Global Context Module (GCM) generates multi-scale features to exploit the rich global context information in order to obtain the correlation between different salient regions and mitigate the dilution of high-level features. Experimental results on the latest light field dataset show that the textual method outperforms the compared methods both quantitatively and qualitatively, and is able to detect accurately complete salient objects and obtain clear saliency maps from similar front/background scenes.
2023, 45(12): 4499-4508.
doi: 10.11999/JEIT221172
Abstract:
In order to improve the accuracy and convergence speed of prediction, a combined prediction model, based on Complementary Ensemble Empirical Mode Decomposition (CEEMD) and Bidirectional Gated Recurrent Unit (BiGRU) optimized by Improved Sparrow Search Algorithm (ISSA), is proposed to deal with the nonlinear, non-stationary and temporal correlation of short-term traffic flow prediction. Firstly, considering the end-point flying wing problem, the traffic flow sequence is decomposed into Intrinsic Mode Function (IMF) components that reflect the trend, periodicity and randomness of road network traffic by improved CEEMD algorithm, which extracts effectively the prior features; Then, the BiGRU network is used to mine the temporal correlation in traffic flow sequence. To fear the local optimum, and improve the global search and local mining ability of Sparrow Search Algorithm (SSA), ISSA is used instead of gradient descent method to iterate the BiGRU network weights. The ablation experiment results show that each component in the combined prediction model plays a positive role in improving the prediction accuracy. The prediction performance under different traffic flow sets is better than the compared algorithm, showing accurate and fast prediction performance with good generalization ability.
In order to improve the accuracy and convergence speed of prediction, a combined prediction model, based on Complementary Ensemble Empirical Mode Decomposition (CEEMD) and Bidirectional Gated Recurrent Unit (BiGRU) optimized by Improved Sparrow Search Algorithm (ISSA), is proposed to deal with the nonlinear, non-stationary and temporal correlation of short-term traffic flow prediction. Firstly, considering the end-point flying wing problem, the traffic flow sequence is decomposed into Intrinsic Mode Function (IMF) components that reflect the trend, periodicity and randomness of road network traffic by improved CEEMD algorithm, which extracts effectively the prior features; Then, the BiGRU network is used to mine the temporal correlation in traffic flow sequence. To fear the local optimum, and improve the global search and local mining ability of Sparrow Search Algorithm (SSA), ISSA is used instead of gradient descent method to iterate the BiGRU network weights. The ablation experiment results show that each component in the combined prediction model plays a positive role in improving the prediction accuracy. The prediction performance under different traffic flow sets is better than the compared algorithm, showing accurate and fast prediction performance with good generalization ability.
2023, 45(12): 4509-4518.
doi: 10.11999/JEIT221569
Abstract:
In view of the problem of synchronous authentication of wireless target identification in the application environment of one-to-many notification relationship entity authentication and key agreement, a double key combination verification theorem is designed. The theorem of interactive dynamic authentication and working key synchronization uopdate is proposed and proved. Based on the dynamic key matching rule of trusted identity, a combinatorial verification authentication model of key synchronization update is constructed. A kind of key synchronous updating combinatorial verification authentication method is proposed. The design criteria of wireless target identification protocol, such as double key combination verification, moderate message retransmission, reasonable simulation of analog channel signal-to-noise ratio, are given. It breaks through the key technology of synchronization authentication in the wireless target identification protocol. The problem of dynamic authentication of entity identity and synchronous updating of working key in entity authentication and key agreement is solved. Taking a class of wireless target identification protocols as an example, the application of this kind of methods is analyzed and illustrated. The formal proof of the protocol is given by a class of constructive attack methods based on strand space theory, and the actual security of the protocol is analyzed by conventional attack methods. Compared with other design methods of synchronous authentication for interactive cryptographic protocols, this method has dynamic authentication. The synchronous authentication scheme designed by this method has the advantages of high security, less computation and only one iteration, which can be applied to wireless target identification in large-scale and complex environment.
In view of the problem of synchronous authentication of wireless target identification in the application environment of one-to-many notification relationship entity authentication and key agreement, a double key combination verification theorem is designed. The theorem of interactive dynamic authentication and working key synchronization uopdate is proposed and proved. Based on the dynamic key matching rule of trusted identity, a combinatorial verification authentication model of key synchronization update is constructed. A kind of key synchronous updating combinatorial verification authentication method is proposed. The design criteria of wireless target identification protocol, such as double key combination verification, moderate message retransmission, reasonable simulation of analog channel signal-to-noise ratio, are given. It breaks through the key technology of synchronization authentication in the wireless target identification protocol. The problem of dynamic authentication of entity identity and synchronous updating of working key in entity authentication and key agreement is solved. Taking a class of wireless target identification protocols as an example, the application of this kind of methods is analyzed and illustrated. The formal proof of the protocol is given by a class of constructive attack methods based on strand space theory, and the actual security of the protocol is analyzed by conventional attack methods. Compared with other design methods of synchronous authentication for interactive cryptographic protocols, this method has dynamic authentication. The synchronous authentication scheme designed by this method has the advantages of high security, less computation and only one iteration, which can be applied to wireless target identification in large-scale and complex environment.
2023, 45(12): 4519-4529.
doi: 10.11999/JEIT221332
Abstract:
With the development of modern science and technology, people have higher and higher requirements for the security of image information transmission, and the image encryption scheme based on chaos theory has attracted more and more attention. In this paper, a novel optical chaotic image encryption transmission system and a “self-encryption” algorithm for images are proposed and demonstrated. The Master Laser (ML) of the system is injected into the other three Semiconductor Lasers (SLs) respectively after full-optical feedback, then three synchronous chaotic sequences are generated. Before encrypting the image, the plaintext image is preprocessed, and two images are obtained, one is the image after the quotient of the plain image, the other is the image after the redundancy of the plain image. The chaotic sequence of the sender is used to encrypt, steganograph and spread the two preprocessed images for many times, and then the ciphertext image is obtained. The experimental results show that the pixel values of the ciphertext images obtained in this paper are evenly distributed, the correlation between each pixel is broken, and both NPCR and UACI are close to the ideal value. The image preprocessing method proposed in this paper can effectively make the image pixel value more concentrated, more uniform distribution. Combining with the optical chaos to encrypt the image, it greatly improves the security of the transmitted image.
With the development of modern science and technology, people have higher and higher requirements for the security of image information transmission, and the image encryption scheme based on chaos theory has attracted more and more attention. In this paper, a novel optical chaotic image encryption transmission system and a “self-encryption” algorithm for images are proposed and demonstrated. The Master Laser (ML) of the system is injected into the other three Semiconductor Lasers (SLs) respectively after full-optical feedback, then three synchronous chaotic sequences are generated. Before encrypting the image, the plaintext image is preprocessed, and two images are obtained, one is the image after the quotient of the plain image, the other is the image after the redundancy of the plain image. The chaotic sequence of the sender is used to encrypt, steganograph and spread the two preprocessed images for many times, and then the ciphertext image is obtained. The experimental results show that the pixel values of the ciphertext images obtained in this paper are evenly distributed, the correlation between each pixel is broken, and both NPCR and UACI are close to the ideal value. The image preprocessing method proposed in this paper can effectively make the image pixel value more concentrated, more uniform distribution. Combining with the optical chaos to encrypt the image, it greatly improves the security of the transmitted image.
2023, 45(12): 4530-4538.
doi: 10.11999/JEIT221164
Abstract:
The energy leakage in the decapsulation process of lattice-based cryptography is analyzed and a message recovery method targeting the message decoding with profiling and ciphertexts rotation is proposed in this paper. The templates are constructed using Hamming weight model as well as Normalized Inter-Class Variance (NICV) for the intermediate state of decoded bytes. The special ciphertexts are built by rotating the original ciphertexts. Combining the energy leakage generated during the calculations, the secret messages and shared keys are recovered. Experiments and tests are carried out with Saber and its variants on the ChipWhisperer-STM32F303 board and the results indicate that the proposed method can successfully recover the secret message and shared key of the encapsulation stage. It only needs 900 energy traces to complete the construction for templates and a total of 32 power traces in recovering the secret message. The success rate of message recovery reaches 66.7% under the condition of no increasing the Signal-to-Noise Ratio (SNR), and 98.43% under the condition of sufficient SNR.
The energy leakage in the decapsulation process of lattice-based cryptography is analyzed and a message recovery method targeting the message decoding with profiling and ciphertexts rotation is proposed in this paper. The templates are constructed using Hamming weight model as well as Normalized Inter-Class Variance (NICV) for the intermediate state of decoded bytes. The special ciphertexts are built by rotating the original ciphertexts. Combining the energy leakage generated during the calculations, the secret messages and shared keys are recovered. Experiments and tests are carried out with Saber and its variants on the ChipWhisperer-STM32F303 board and the results indicate that the proposed method can successfully recover the secret message and shared key of the encapsulation stage. It only needs 900 energy traces to complete the construction for templates and a total of 32 power traces in recovering the secret message. The success rate of message recovery reaches 66.7% under the condition of no increasing the Signal-to-Noise Ratio (SNR), and 98.43% under the condition of sufficient SNR.
2023, 45(12): 4539-4547.
doi: 10.11999/JEIT221296
Abstract:
Focusing on the problems of the traditional network traffic anomaly detection methods, such as low recognition accuracy, weak representation ability, poor generalization ability, and ignoring the relationship between features, a network traffic anomaly detection method based on quadratic feature extraction and BiLSTM-Attention is proposed. By using the Bidirectional Long Short-Term Memory network (BiLSTM) to learn the feature relationship between the data, the feature of the data is extracted, on this basis, a feature importance weight evaluation rule based on attention mechanism is defined, and the feature vector generated by BiLSTM is given corresponding weight according to the feature importance to complete the secondary feature extraction of data. Finally, a design idea of “total score first and then subdivision” is proposed to construct a network traffic anomaly detection model to implement anomaly detection of multi-classified network traffic. The experimental results show that the method proposed in this paper is better than the traditional single model in performance, and has good representation ability and generalization ability.
Focusing on the problems of the traditional network traffic anomaly detection methods, such as low recognition accuracy, weak representation ability, poor generalization ability, and ignoring the relationship between features, a network traffic anomaly detection method based on quadratic feature extraction and BiLSTM-Attention is proposed. By using the Bidirectional Long Short-Term Memory network (BiLSTM) to learn the feature relationship between the data, the feature of the data is extracted, on this basis, a feature importance weight evaluation rule based on attention mechanism is defined, and the feature vector generated by BiLSTM is given corresponding weight according to the feature importance to complete the secondary feature extraction of data. Finally, a design idea of “total score first and then subdivision” is proposed to construct a network traffic anomaly detection model to implement anomaly detection of multi-classified network traffic. The experimental results show that the method proposed in this paper is better than the traditional single model in performance, and has good representation ability and generalization ability.
2023, 45(12): 4548-4555.
doi: 10.11999/JEIT221414
Abstract:
Deep Neural Networks (DNN) are widely used due to their excellent performance, but the problem of being vulnerable to adversarial examples makes them face huge security risks.Through the visualization of the convolution process of DNN, it is found that with the deepening of the convolution layers, the disturbance of the original input caused by the adversarial attack becomes more obvious. Based on this finding, a defense algorithm based on Momentum Enhanced Feature maps (MEF) is proposed by adopting the idea of revising the backward results by the forward results in the momentum method. The MEF algorithm deploys a feature enhancement layer on the convolutional layer of the DNN to form a Feature Enhancement Block (FEB). The FEB combines the original input and the feature map of the shallow convolutional layer to generate a feature enhancement map, and then uses the feature enhancement map to enhance the deep features map. While, in order to ensure the effectiveness of the feature enhancement map of each layer, the enhanced feature map will further update the feature enhancement map. In order to verify the effectiveness of the MEF algorithm, various white-box and black-box attacks are used to attack the DNN model deployed with the MEF algorithm, the results show that in the Project Gradient Descent (PGD) and Fast Gradient Sign Method (FGSM) attack experiment, the recognition accuracy of MEF algorithm for adversarial samples is 3%～5% higher than that of Adversarial Training (AT), and the recognition accuracy of clean samples is also improved. Furthermore, when tested with stronger adversarial attack methods than training, the MEF algorithm exhibits stronger robustness compared with the currently advanced Parametric Noise Injection algorithm (PNI) and Learn2Perturb algorithm (L2P).
Deep Neural Networks (DNN) are widely used due to their excellent performance, but the problem of being vulnerable to adversarial examples makes them face huge security risks.Through the visualization of the convolution process of DNN, it is found that with the deepening of the convolution layers, the disturbance of the original input caused by the adversarial attack becomes more obvious. Based on this finding, a defense algorithm based on Momentum Enhanced Feature maps (MEF) is proposed by adopting the idea of revising the backward results by the forward results in the momentum method. The MEF algorithm deploys a feature enhancement layer on the convolutional layer of the DNN to form a Feature Enhancement Block (FEB). The FEB combines the original input and the feature map of the shallow convolutional layer to generate a feature enhancement map, and then uses the feature enhancement map to enhance the deep features map. While, in order to ensure the effectiveness of the feature enhancement map of each layer, the enhanced feature map will further update the feature enhancement map. In order to verify the effectiveness of the MEF algorithm, various white-box and black-box attacks are used to attack the DNN model deployed with the MEF algorithm, the results show that in the Project Gradient Descent (PGD) and Fast Gradient Sign Method (FGSM) attack experiment, the recognition accuracy of MEF algorithm for adversarial samples is 3%～5% higher than that of Adversarial Training (AT), and the recognition accuracy of clean samples is also improved. Furthermore, when tested with stronger adversarial attack methods than training, the MEF algorithm exhibits stronger robustness compared with the currently advanced Parametric Noise Injection algorithm (PNI) and Learn2Perturb algorithm (L2P).
2023, 45(12): 4556-4565.
doi: 10.11999/JEIT221083
Abstract:
In recent years, nonlinear dynamics problems based on memristors have received much attention. In this paper, binary-valued and tri-valued memristors are used as examples to analyze the influence of binary-valued and multi-value memristors on the dynamic characteristics of chaotic systems. Firstly, the binary-valued memristor is introduced into the Chen system, and a four-dimensional Binary-valued Memristor-based Chaotic System(BMCS) is constructed. Secondly, a tri-valued memristor is used to replace the binary-valued memristor in the above system, and a four-dimensional Tri-valued Memristor-based Chaotic System(TMCS) is constructed. Through theoretical analysis and numerical simulation, the differences of dynamic characteristics between the two chaotic systems are compared from multiple perspectives, such as Lyapunov exponent, bifurcation diagram, equilibrium point of the system, system stability, sensitivity to initial value and system complexity analysis, etc. The results show that the two memristor-based chaotic systems have infinite equilibrium points, the attractors generated by both are hidden attractors, and both have transient chaotic phenomena, but the Tri-valued memristor chaotic system has Hyper-chaos, and has stronger initial value sensitivity than Binary-valued memristor chaotic system. Further, the Tri-valued memristor chaotic system has a larger parameter value interval than the Binary-valued memristor chaotic system to obtain chaotic sequences with sufficiently high complexity. Through analysis, it is concluded that the chaotic system based on Tri-valued memristor can generate more complex dynamic characteristics and more complex chaotic signal than the chaotic system based on binary-valued memristor.
In recent years, nonlinear dynamics problems based on memristors have received much attention. In this paper, binary-valued and tri-valued memristors are used as examples to analyze the influence of binary-valued and multi-value memristors on the dynamic characteristics of chaotic systems. Firstly, the binary-valued memristor is introduced into the Chen system, and a four-dimensional Binary-valued Memristor-based Chaotic System(BMCS) is constructed. Secondly, a tri-valued memristor is used to replace the binary-valued memristor in the above system, and a four-dimensional Tri-valued Memristor-based Chaotic System(TMCS) is constructed. Through theoretical analysis and numerical simulation, the differences of dynamic characteristics between the two chaotic systems are compared from multiple perspectives, such as Lyapunov exponent, bifurcation diagram, equilibrium point of the system, system stability, sensitivity to initial value and system complexity analysis, etc. The results show that the two memristor-based chaotic systems have infinite equilibrium points, the attractors generated by both are hidden attractors, and both have transient chaotic phenomena, but the Tri-valued memristor chaotic system has Hyper-chaos, and has stronger initial value sensitivity than Binary-valued memristor chaotic system. Further, the Tri-valued memristor chaotic system has a larger parameter value interval than the Binary-valued memristor chaotic system to obtain chaotic sequences with sufficiently high complexity. Through analysis, it is concluded that the chaotic system based on Tri-valued memristor can generate more complex dynamic characteristics and more complex chaotic signal than the chaotic system based on binary-valued memristor.
2023, 45(12): 4566-4575.
doi: 10.11999/JEIT221315
Abstract:
OptoElectronic Tweezer (OET) is a micro-scale optical manipulation technology based on photoinduced electrophoretic effect. It can accurately control small targets in the complex environment of fluid field, photoelectric field and biological force field, and has important applications to cell operation, micromechanical system and other fields. Optoelectronic tweezers technology can be used alone or in conjunction with other technologies, and has been widely used. To date, research based on optoelectronic tweezers has focused on manipulation, assembly, and synthesis of micro and nanomaterials; manipulation, isolation, and analysis of individual cells/molecules; analysis and acquisition of cell intrinsic properties; electroporation, fusion, and lysis of cells; preparation of cell-encapsulated biomaterials and biological structures; development of optical fluid devices for fluid transport. These works demonstrate the superior performance and unique versatility and flexibility of the optoelectronic tweezers technology. The existing application of optoelectronic tweezers technology are systematically presented in the paper and the application prospect, limitation and development trend of this technology are summarised.
OptoElectronic Tweezer (OET) is a micro-scale optical manipulation technology based on photoinduced electrophoretic effect. It can accurately control small targets in the complex environment of fluid field, photoelectric field and biological force field, and has important applications to cell operation, micromechanical system and other fields. Optoelectronic tweezers technology can be used alone or in conjunction with other technologies, and has been widely used. To date, research based on optoelectronic tweezers has focused on manipulation, assembly, and synthesis of micro and nanomaterials; manipulation, isolation, and analysis of individual cells/molecules; analysis and acquisition of cell intrinsic properties; electroporation, fusion, and lysis of cells; preparation of cell-encapsulated biomaterials and biological structures; development of optical fluid devices for fluid transport. These works demonstrate the superior performance and unique versatility and flexibility of the optoelectronic tweezers technology. The existing application of optoelectronic tweezers technology are systematically presented in the paper and the application prospect, limitation and development trend of this technology are summarised.
2023, 45(12): 4576-4584.
doi: 10.11999/JEIT221181
Abstract:
The requirement of miniaturization and lightweight of space Travelling Wave Tube (TWT) predistortion circuit makes the circuit debugging more difficult. Therefore, an accurate simulation and design method of predistortion circuit is urgently needed to guide the product design. Based on the analysis of Schottky diode equivalent circuit model, the diode MA4E2039 is selected as a nonlinear generator, and the diode simulation model of MA4E2039 is established. Then, the key parameters affecting the performance of the circuit are obtained by analyzing the structure of the reflective predistortion circuit, and these key parameters are simulated accurately in the co-simulation stage of components and layout. Finally, the pre-distortion circuit processed according to the simulation results is tested, and it is found that the deviation between the simulation results and the measured results is less than 15%. By cascading the predistortion circuit with the K-band TWTA, the third-order intermodulation reaches 23.77 dBc while the IBO=4 dB. Therefore, this method can be used to guide the design of pre-distortion circuit of space TWT, help to improve the product development cycle, and also has important guiding significance for the miniaturization design of pre-distortion circuit.
The requirement of miniaturization and lightweight of space Travelling Wave Tube (TWT) predistortion circuit makes the circuit debugging more difficult. Therefore, an accurate simulation and design method of predistortion circuit is urgently needed to guide the product design. Based on the analysis of Schottky diode equivalent circuit model, the diode MA4E2039 is selected as a nonlinear generator, and the diode simulation model of MA4E2039 is established. Then, the key parameters affecting the performance of the circuit are obtained by analyzing the structure of the reflective predistortion circuit, and these key parameters are simulated accurately in the co-simulation stage of components and layout. Finally, the pre-distortion circuit processed according to the simulation results is tested, and it is found that the deviation between the simulation results and the measured results is less than 15%. By cascading the predistortion circuit with the K-band TWTA, the third-order intermodulation reaches 23.77 dBc while the IBO=4 dB. Therefore, this method can be used to guide the design of pre-distortion circuit of space TWT, help to improve the product development cycle, and also has important guiding significance for the miniaturization design of pre-distortion circuit.
