Email alert
2021 Vol. 43, No. 10
Display Method:
2021, 43(10): 2757-2770.
doi: 10.11999/JEIT210118
Abstract:
Sparse Code Multiple Access (SCMA) is a non-orthogonal multiple access scheme with high spectrum efficiency and massive connectivity. Multi-user detection is a key issue of SCMA system and how to reduce the complexity of detection algorithms has become a research hotspot in the field of multiple access. This paper reviews the existing multi-user detection algorithms from different factors that affect the complexity of the algorithms. Furthermore, the principle and performance of several typical multi-user detection algorithms are introduced and compared. In addition, the improved ideas of multi-user detection algorithms for SCMA system are pointed out in this paper. Finally, the future development trend and challenges of SCMA detection algorithms are summarized and discussed.
Sparse Code Multiple Access (SCMA) is a non-orthogonal multiple access scheme with high spectrum efficiency and massive connectivity. Multi-user detection is a key issue of SCMA system and how to reduce the complexity of detection algorithms has become a research hotspot in the field of multiple access. This paper reviews the existing multi-user detection algorithms from different factors that affect the complexity of the algorithms. Furthermore, the principle and performance of several typical multi-user detection algorithms are introduced and compared. In addition, the improved ideas of multi-user detection algorithms for SCMA system are pointed out in this paper. Finally, the future development trend and challenges of SCMA detection algorithms are summarized and discussed.
2021, 43(10): 2771-2780.
doi: 10.11999/JEIT200451
Abstract:
Sea clutter is one of the main limiting factors influencing the target detection performance of nautical radars, and its physical mechanism at low grazing angles has been widely studied. Due to the difference in the generation mechanism, it is difficult to adopt directly the existing diversified low clutter angle sea clutter model and characteristic conclusions in the cognitive study of the sea clutter characteristics under large grazing angle.The study of sea clutter characteristics under large grazing angles is still extremely complicated and has to be systematic. Based on the commonly used data, this paper reviews and summarizes the research developments in sea clutter at large grazing angles. It concentrates on the properties that are of most interest for target detection at large grazing angles: the characteristics, modeling methods and detection technology algorithms. The main research results are concluded and the development direction of the technology is prospected.
Sea clutter is one of the main limiting factors influencing the target detection performance of nautical radars, and its physical mechanism at low grazing angles has been widely studied. Due to the difference in the generation mechanism, it is difficult to adopt directly the existing diversified low clutter angle sea clutter model and characteristic conclusions in the cognitive study of the sea clutter characteristics under large grazing angle.The study of sea clutter characteristics under large grazing angles is still extremely complicated and has to be systematic. Based on the commonly used data, this paper reviews and summarizes the research developments in sea clutter at large grazing angles. It concentrates on the properties that are of most interest for target detection at large grazing angles: the characteristics, modeling methods and detection technology algorithms. The main research results are concluded and the development direction of the technology is prospected.
2021, 43(10): 2781-2789.
doi: 10.11999/JEIT200777
Abstract:
In spaceborne Synthetic Aperture Radar (SAR) imaging, the coherence of azimuth signals might be degraded by spatio-temporal variations of the background ionosphere, especially for low-frequency systems. In this paper, the azimuth temporal-varying Slant Total Electron Content (STEC) is attributed to three factors: the spatial- and temporal-varying Vertical Total Electron Content (VTEC), and the propagation path variation. Derivative of each order of the azimuth-varying STEC is analyzed as to each factor. A uniform analytical model is established, that is the third-order Taylor expansion model of SAR azimuth signals influenced by the temporal-varying STEC. The analytical expressions of the azimuth shift and phase errors are derived. Based on this model, thresholds of the varying STEC derivatives are derived for different spaceborne SAR systems. Signal-level simulations are performed by means of the VTEC real data and the International Reference Ionosphere (IRI) model. Numerical analyses and signal-level simulations indicate that the spatial-varying VTEC and the propagation path variation highlight in the low-orbit P-band SAR system, while the temporal-varying VTEC becomes a predominant factor that results in the azimuth temporal-varying STEC in the medium- or high-orbit SAR system. As the carrier frequency decreases and the synthetic aperture time increases, the azimuth imaging performance is more susceptible to the azimuth temporal-varying STEC.
In spaceborne Synthetic Aperture Radar (SAR) imaging, the coherence of azimuth signals might be degraded by spatio-temporal variations of the background ionosphere, especially for low-frequency systems. In this paper, the azimuth temporal-varying Slant Total Electron Content (STEC) is attributed to three factors: the spatial- and temporal-varying Vertical Total Electron Content (VTEC), and the propagation path variation. Derivative of each order of the azimuth-varying STEC is analyzed as to each factor. A uniform analytical model is established, that is the third-order Taylor expansion model of SAR azimuth signals influenced by the temporal-varying STEC. The analytical expressions of the azimuth shift and phase errors are derived. Based on this model, thresholds of the varying STEC derivatives are derived for different spaceborne SAR systems. Signal-level simulations are performed by means of the VTEC real data and the International Reference Ionosphere (IRI) model. Numerical analyses and signal-level simulations indicate that the spatial-varying VTEC and the propagation path variation highlight in the low-orbit P-band SAR system, while the temporal-varying VTEC becomes a predominant factor that results in the azimuth temporal-varying STEC in the medium- or high-orbit SAR system. As the carrier frequency decreases and the synthetic aperture time increases, the azimuth imaging performance is more susceptible to the azimuth temporal-varying STEC.
2021, 43(10): 2790-2797.
doi: 10.11999/JEIT200981
Abstract:
To avoid the detection to own fighters from adversary’s radar, a method is proposed in this paper, under which the fighter transmits radio frequency shielding signals. If Constant False Alarm Rate (CFAR) detector is used in adversary’s radars, adversary’s noise level of radar channel is overestimated when the fighter transmits the certain envelope radio frequency shielding signals by own fighters. The effect of making own fighters invisible to the radar is achieved while radio frequency shielding signals avoid being intercepted by adversary electronic reconnaissance equipment. For these purposes, it is needed to estimate the time of arrival of the adversary’s radar signal to the fighters, evaluate the enemy’s detection capability, and design the envelope and power of the shielding noise. Based on the Cell Average-Constant False Alarm Rate (CA-CFRA) detector, numerical simulation is carried out to analyze the relationship between the radio frequency shielding envelope and the interception probability of the shielding signal and the detection probability of the target. These simulation results show that the shielding method can achieve a better radio frequency shielding effect with properly designed shielding signal envelope.
To avoid the detection to own fighters from adversary’s radar, a method is proposed in this paper, under which the fighter transmits radio frequency shielding signals. If Constant False Alarm Rate (CFAR) detector is used in adversary’s radars, adversary’s noise level of radar channel is overestimated when the fighter transmits the certain envelope radio frequency shielding signals by own fighters. The effect of making own fighters invisible to the radar is achieved while radio frequency shielding signals avoid being intercepted by adversary electronic reconnaissance equipment. For these purposes, it is needed to estimate the time of arrival of the adversary’s radar signal to the fighters, evaluate the enemy’s detection capability, and design the envelope and power of the shielding noise. Based on the Cell Average-Constant False Alarm Rate (CA-CFRA) detector, numerical simulation is carried out to analyze the relationship between the radio frequency shielding envelope and the interception probability of the shielding signal and the detection probability of the target. These simulation results show that the shielding method can achieve a better radio frequency shielding effect with properly designed shielding signal envelope.
2021, 43(10): 2798-2804.
doi: 10.11999/JEIT200973
Abstract:
An algorithm based on high-efficiency FRactional Fourier Transform (FRFT) and fourth-order origin moments in the fractional-domain spectrum is proposed to estimate quickly the chirp signal at low signal-to-noise ratio. Firstly, the initial interval of the rotation order is determined by the sign of the FM slope. Then, the rotation order is estimated roughly by the efficient FRFT algorithm. Finally, the search interval and step size are determined according to the fourth-order origin moments of the spectrum in the fractional-domain. The simulation results show that the Linear Frequency Modulation (LFM) signal can be detected under low signal-to-noise ratio and the parameters of the signal can be estimated accurately using this algorithm, and the signal can be detected with a small amount of calculation.
An algorithm based on high-efficiency FRactional Fourier Transform (FRFT) and fourth-order origin moments in the fractional-domain spectrum is proposed to estimate quickly the chirp signal at low signal-to-noise ratio. Firstly, the initial interval of the rotation order is determined by the sign of the FM slope. Then, the rotation order is estimated roughly by the efficient FRFT algorithm. Finally, the search interval and step size are determined according to the fourth-order origin moments of the spectrum in the fractional-domain. The simulation results show that the Linear Frequency Modulation (LFM) signal can be detected under low signal-to-noise ratio and the parameters of the signal can be estimated accurately using this algorithm, and the signal can be detected with a small amount of calculation.
2021, 43(10): 2805-2814.
doi: 10.11999/JEIT200975
Abstract:
According to the low sampling efficiency of particle filter track before detecting in high dimension state space with bearing-frequency measurements of passive sonar, a two-hierarchy sampling method based on the observability of leg-by-leg maneuver is proposed. Firstly, the observability of leg-by-leg maneuver is analyzed. Secondly, the target motion model in polar coordinate system is build. Based on the uniform distribution of the distance and normal velocity of particles relative to the observation station, the method of mapping the target state vector in polar coordinate system to rectangular coordinate system is proposed. Finally, in order to improve the convergence of the filter, the covariance matrix of process noise is adaptively adjusted according to the spatial distribution of particle. Simulation results show that, compared with the traditional method, the proposed method can increase the filter convergence rate by about 47.6%, reduce the distance estimation error by about 329 m and reduce the convergence time by about 450 s.
According to the low sampling efficiency of particle filter track before detecting in high dimension state space with bearing-frequency measurements of passive sonar, a two-hierarchy sampling method based on the observability of leg-by-leg maneuver is proposed. Firstly, the observability of leg-by-leg maneuver is analyzed. Secondly, the target motion model in polar coordinate system is build. Based on the uniform distribution of the distance and normal velocity of particles relative to the observation station, the method of mapping the target state vector in polar coordinate system to rectangular coordinate system is proposed. Finally, in order to improve the convergence of the filter, the covariance matrix of process noise is adaptively adjusted according to the spatial distribution of particle. Simulation results show that, compared with the traditional method, the proposed method can increase the filter convergence rate by about 47.6%, reduce the distance estimation error by about 329 m and reduce the convergence time by about 450 s.
2021, 43(10): 2815-2823.
doi: 10.11999/JEIT210234
Abstract:
Detection and tracking of low signal-to-noise ratio nonlinear frequency modulated signal can be effectively solved by Track-Before-Detecting (TBD) algorithms based on particle filters. However, the algorithms are high in computational complexity and hard to be implemented in parallel. Furthermore, because of the comparatively long convergence processing, the detection and state estimation capabilities of the particle filters based methods are limited. In this paper, a cost-reference particle filter bank is proposed, which does not depend on the distribution of the system and has an entirely parallel structure. Then a detection method based on the cost-reference particle filter bank is proposed. Simulation results of two nonlinear frequency modulated signals detection and estimation illustrate that the propose method has better performance in detection, estimation, and running speed than similar methods, such as particle filter based track-before-detecting algorithm, Rutten particle filter based TBD algorithm.
Detection and tracking of low signal-to-noise ratio nonlinear frequency modulated signal can be effectively solved by Track-Before-Detecting (TBD) algorithms based on particle filters. However, the algorithms are high in computational complexity and hard to be implemented in parallel. Furthermore, because of the comparatively long convergence processing, the detection and state estimation capabilities of the particle filters based methods are limited. In this paper, a cost-reference particle filter bank is proposed, which does not depend on the distribution of the system and has an entirely parallel structure. Then a detection method based on the cost-reference particle filter bank is proposed. Simulation results of two nonlinear frequency modulated signals detection and estimation illustrate that the propose method has better performance in detection, estimation, and running speed than similar methods, such as particle filter based track-before-detecting algorithm, Rutten particle filter based TBD algorithm.
2021, 43(10): 2824-2831.
doi: 10.11999/JEIT200748
Abstract:
The order of traditional Spectrum Spread and Compression (SSC) blind frequency shift jamming is an integer. To achieve accurate position interference, different processing delays need to be adjusted, which has certain limitations in practical application. In this paper, the integer order blind frequency shift technology is improved, and a Linear Frequency Modulation (LFM) radar jamming technology based on non-integer order SSC blind frequency shift is proposed. The jamming can be realized by changing the precise position of the radar in different situations. In this paper, an efficient implementation method of blind frequency shift jamming for non-integer order SSC is derived. Meanwhile, the initial phase of the signal is controlled by the Newman sequence to reduce the Peak-to-Average Power Ratio (PAPR) of the jamming signal. The simulation results show that the algorithm can achieve false target deception jamming and coherent dense suppression jamming under the processing delay of a specific jammer, which can effectively counter pulse compression radar, and has good engineering application value.
The order of traditional Spectrum Spread and Compression (SSC) blind frequency shift jamming is an integer. To achieve accurate position interference, different processing delays need to be adjusted, which has certain limitations in practical application. In this paper, the integer order blind frequency shift technology is improved, and a Linear Frequency Modulation (LFM) radar jamming technology based on non-integer order SSC blind frequency shift is proposed. The jamming can be realized by changing the precise position of the radar in different situations. In this paper, an efficient implementation method of blind frequency shift jamming for non-integer order SSC is derived. Meanwhile, the initial phase of the signal is controlled by the Newman sequence to reduce the Peak-to-Average Power Ratio (PAPR) of the jamming signal. The simulation results show that the algorithm can achieve false target deception jamming and coherent dense suppression jamming under the processing delay of a specific jammer, which can effectively counter pulse compression radar, and has good engineering application value.
2021, 43(10): 2832-2839.
doi: 10.11999/JEIT200604
Abstract:
The compact High-Frequency Surface Wave Radar (HFSWR) has low spatial resolution for target detection due to its reduced aperture size of the receiving antenna array. The sequential plot-to-track association method used in multi-target tracking algorithms is prone to erroneous association, which easily leads to track fragmentation and false tracking. In order to solve this problem, regarding the multi-target plot-to-track association as an optimal allocation problem, an optimal multi-target plot-to-track association method based on JVC (Jonker-Volgenant-Castanon) algorithm is proposed. For multiple tracks with common candidate plots in their overlapped association gate, firstly, the similarity between their candidate plots and all tracks is calculated using the minimal cost function with target Doppler velocity, range and azimuth as parameters and an association cost matrix is formed. Then, the optimal association result is achieved by minimizing the total association cost using the JVC algorithm. Both simulation and field target data are used to carry out the plot-to-track association experiment, and the association results are compared with those of the sequential nearest neighbor association method. The experimental results show that the track length obtained by the proposed method is superior to that of the sequential nearest neighbor method, thus the track continuity is improved.
The compact High-Frequency Surface Wave Radar (HFSWR) has low spatial resolution for target detection due to its reduced aperture size of the receiving antenna array. The sequential plot-to-track association method used in multi-target tracking algorithms is prone to erroneous association, which easily leads to track fragmentation and false tracking. In order to solve this problem, regarding the multi-target plot-to-track association as an optimal allocation problem, an optimal multi-target plot-to-track association method based on JVC (Jonker-Volgenant-Castanon) algorithm is proposed. For multiple tracks with common candidate plots in their overlapped association gate, firstly, the similarity between their candidate plots and all tracks is calculated using the minimal cost function with target Doppler velocity, range and azimuth as parameters and an association cost matrix is formed. Then, the optimal association result is achieved by minimizing the total association cost using the JVC algorithm. Both simulation and field target data are used to carry out the plot-to-track association experiment, and the association results are compared with those of the sequential nearest neighbor association method. The experimental results show that the track length obtained by the proposed method is superior to that of the sequential nearest neighbor method, thus the track continuity is improved.
2021, 43(10): 2840-2847.
doi: 10.11999/JEIT210620
Abstract:
Different from the traditional multi-target tracking problem which has the measurements to targets data association uncertainty problem, the multistatic passive radar multi-target tacking system has the additional measurements to illuminators data association uncertainty problem, which means the data association relationship is three dimensional. A novel target-measurement-illuminator Probabilistic Multiple Hypothesis Tracking (PMHT) algorithm is proposed, which introduces a new data association variable to represent the data association relationship. The proposed algorithm is based on the Expectation-Maximization (EM). To handle the nonlinear problem of range-Doppler measurements, the Unscented Kalman Smoother (UKS) is used to get the multi-targets’ estimated states. To increase the data association accuracy, the measurements are set to mixture Gaussian distribution. Simulation results show that for the FKIE passive radar data set, the proposed algorithm can track multi-targets effectively in dense clutter environment.
Different from the traditional multi-target tracking problem which has the measurements to targets data association uncertainty problem, the multistatic passive radar multi-target tacking system has the additional measurements to illuminators data association uncertainty problem, which means the data association relationship is three dimensional. A novel target-measurement-illuminator Probabilistic Multiple Hypothesis Tracking (PMHT) algorithm is proposed, which introduces a new data association variable to represent the data association relationship. The proposed algorithm is based on the Expectation-Maximization (EM). To handle the nonlinear problem of range-Doppler measurements, the Unscented Kalman Smoother (UKS) is used to get the multi-targets’ estimated states. To increase the data association accuracy, the measurements are set to mixture Gaussian distribution. Simulation results show that for the FKIE passive radar data set, the proposed algorithm can track multi-targets effectively in dense clutter environment.
2021, 43(10): 2848-2853.
doi: 10.11999/JEIT200559
Abstract:
To solve the problem that the carrier antenna beam directs to Beidou satellite in Beidou short-message communication, a direction algorithm of the antenna beam is proposed. According to the position and the attitude of the carrier, the angle of the antenna beam directing to Beidou satellite is calculated. Based on the synthetic beam gain of the carrier antenna and the beam gain of the Beidou satellite, the optimal satellite is selected and directed to. This algorithm can ensure that the carrier antenna beam directs to the optimal satellite in real time when the carrier is moving, and can greatly improve the performance of Beidou short-message communication.
To solve the problem that the carrier antenna beam directs to Beidou satellite in Beidou short-message communication, a direction algorithm of the antenna beam is proposed. According to the position and the attitude of the carrier, the angle of the antenna beam directing to Beidou satellite is calculated. Based on the synthetic beam gain of the carrier antenna and the beam gain of the Beidou satellite, the optimal satellite is selected and directed to. This algorithm can ensure that the carrier antenna beam directs to the optimal satellite in real time when the carrier is moving, and can greatly improve the performance of Beidou short-message communication.
2021, 43(10): 2854-2862.
doi: 10.11999/JEIT201042
Abstract:
In view of the problem that the efficient detection model SSD-MV2 (Single Shot Detector MobileNet V2) has low detection accuracy to underwater multi-scale targets in Synthetic Aperture Sonar (SAS) images, a novel feature extraction module Extended Selective Kernel (ESK) is proposed in this paper. ESK has the advantages of channel scalability, channel selection and few model parameters. At the same time, the basic network and additional feature extraction network of SSD are redesigned by using ESK module, which is named SSD-MV2ESK, and a set of reasonable expansion coefficient and multi-scale coefficient are selected for SSD-MV2ESK. On SST-DET, the mAP of SSD-MV2ESK is 4.71% higher than that of SSD-MV2 when the model parameters are basically the same. The experimental results show that SSD-MV2ESK is suitable for SAR underwater multi-scale target detection task in embedded platform.
In view of the problem that the efficient detection model SSD-MV2 (Single Shot Detector MobileNet V2) has low detection accuracy to underwater multi-scale targets in Synthetic Aperture Sonar (SAS) images, a novel feature extraction module Extended Selective Kernel (ESK) is proposed in this paper. ESK has the advantages of channel scalability, channel selection and few model parameters. At the same time, the basic network and additional feature extraction network of SSD are redesigned by using ESK module, which is named SSD-MV2ESK, and a set of reasonable expansion coefficient and multi-scale coefficient are selected for SSD-MV2ESK. On SST-DET, the mAP of SSD-MV2ESK is 4.71% higher than that of SSD-MV2 when the model parameters are basically the same. The experimental results show that SSD-MV2ESK is suitable for SAR underwater multi-scale target detection task in embedded platform.
2021, 43(10): 2863-2870.
doi: 10.11999/JEIT200697
Abstract:
Focusing on the problem of rather large estimation error in Direction Of Arrival (DOA) estimation caused by grid mismatch in the sparse representation model, an Off-Grid DOA estimation method based on Covariance Matrix Reconstruction (OGCMR) is proposed. Firstly, the offset between the DOA and the grid points is incorporated into the constructed spatial discrete sparse representation model of the received data; After that, based on the reconstructed signal covariance matrix, a sparse representation convex optimization problem associated with DOA estimation can be established; Subsequently, a sampling covariance matrix estimation error convex model is constructed, and then this convex set can be explicitly included into the sparse representation model to improve the performance of sparse signal reconstruction; Finally, an alternating optimization method can be exploited to solve the resultant joint optimization problem to acquire the grid offset parameters as well as the off-grid DOA estimation. Numerical simulations show that, compared with the traditional conventional MUltiple SIgnal Classification(MUSIC), L1-SVD, Sparse and Low-Rank Decomposition based Robust MVDR (SLRD-RMVDR) algorithms and so on, the proposed algorithm has rather better angular resolution and higher DOA estimation accuracy.
Focusing on the problem of rather large estimation error in Direction Of Arrival (DOA) estimation caused by grid mismatch in the sparse representation model, an Off-Grid DOA estimation method based on Covariance Matrix Reconstruction (OGCMR) is proposed. Firstly, the offset between the DOA and the grid points is incorporated into the constructed spatial discrete sparse representation model of the received data; After that, based on the reconstructed signal covariance matrix, a sparse representation convex optimization problem associated with DOA estimation can be established; Subsequently, a sampling covariance matrix estimation error convex model is constructed, and then this convex set can be explicitly included into the sparse representation model to improve the performance of sparse signal reconstruction; Finally, an alternating optimization method can be exploited to solve the resultant joint optimization problem to acquire the grid offset parameters as well as the off-grid DOA estimation. Numerical simulations show that, compared with the traditional conventional MUltiple SIgnal Classification(MUSIC), L1-SVD, Sparse and Low-Rank Decomposition based Robust MVDR (SLRD-RMVDR) algorithms and so on, the proposed algorithm has rather better angular resolution and higher DOA estimation accuracy.
2021, 43(10): 2871-2877.
doi: 10.11999/JEIT200526
Abstract:
In case of the traditional three-satellite time-frequency difference positioning system with unknown elevation, it will produce positioning errors caused by the target speed to the moving state radiation source. A new method of fusion passive tracking is proposed, which uses the Direction Of Arrival (DOA) information of the main star and combines the Time Difference Of Arrival (TDOA) and Frequency Difference Of Arrival (FDOA) information of the three satellites. First, a positioning model is established on the basis of coordinate system transformation, and on this basis, an improved Iterative Extended Kalman Filter (IEKF) algorithm is used to track moving objects of unknown elevation. The simulation results of the algorithm show that this method can realize the positioning, tracking and speed measurement of the moving radiation source without knowing the height information of the target, and the position and speed estimation of the target are improved.
In case of the traditional three-satellite time-frequency difference positioning system with unknown elevation, it will produce positioning errors caused by the target speed to the moving state radiation source. A new method of fusion passive tracking is proposed, which uses the Direction Of Arrival (DOA) information of the main star and combines the Time Difference Of Arrival (TDOA) and Frequency Difference Of Arrival (FDOA) information of the three satellites. First, a positioning model is established on the basis of coordinate system transformation, and on this basis, an improved Iterative Extended Kalman Filter (IEKF) algorithm is used to track moving objects of unknown elevation. The simulation results of the algorithm show that this method can realize the positioning, tracking and speed measurement of the moving radiation source without knowing the height information of the target, and the position and speed estimation of the target are improved.
2021, 43(10): 2878-2885.
doi: 10.11999/JEIT200683
Abstract:
In view of the problem that dual-dispersion channels will reduce the reliability of channel estimation in high-speed mobile environments, a channel estimation algorithm based on compressed sensing is proposed in the input-output model of Orthogonal Time-Frequency-Space (OTFS) modulation system. The maximum Doppler shift and the maximum delay in the channel are employed to determine the size of the pilot transmission matrix in the algorithm. Compared with the traditional Orthogonal Matching Pursuit (OMP) channel estimation algorithms, the pilot resources can be saved in the proposed algorithm while the accuracy of similar channel estimation is guaranteed. Furthermore, the phase rotation of the OTFS modulation symbols is used to improve the rank of the differential matrix. Theoretical analysis and simulation results show that the diversity order of the OTFS system is improved and noise interference is reduced.
In view of the problem that dual-dispersion channels will reduce the reliability of channel estimation in high-speed mobile environments, a channel estimation algorithm based on compressed sensing is proposed in the input-output model of Orthogonal Time-Frequency-Space (OTFS) modulation system. The maximum Doppler shift and the maximum delay in the channel are employed to determine the size of the pilot transmission matrix in the algorithm. Compared with the traditional Orthogonal Matching Pursuit (OMP) channel estimation algorithms, the pilot resources can be saved in the proposed algorithm while the accuracy of similar channel estimation is guaranteed. Furthermore, the phase rotation of the OTFS modulation symbols is used to improve the rank of the differential matrix. Theoretical analysis and simulation results show that the diversity order of the OTFS system is improved and noise interference is reduced.
2021, 43(10): 2886-2893.
doi: 10.11999/JEIT200961
Abstract:
Large-scale disaster events might critically threaten and damage the infrastructure of telecom networks. In view of the virtual network survivability issue under large-scale disaster risks, a Multi-virtual-machine Rapid Cooperative Evacuation (MRCE) mechanism is proposed. The proposed mechanism employs the post-copy technique for live migration of virtual machines. By means of the basic migration bandwidth assignment and upgradation, a rapid cooperative evacuation is conducted on multiple risk virtual machines belonging to a single virtual network, so as to shorten the evacuation time of that virtual network and reduce its damage risk. Simulation results show that the proposed mechanism can achieve better performance than its counterparts in terms of virtual network evacuation completion ratio and average evacuation completion time in different observation periods.
Large-scale disaster events might critically threaten and damage the infrastructure of telecom networks. In view of the virtual network survivability issue under large-scale disaster risks, a Multi-virtual-machine Rapid Cooperative Evacuation (MRCE) mechanism is proposed. The proposed mechanism employs the post-copy technique for live migration of virtual machines. By means of the basic migration bandwidth assignment and upgradation, a rapid cooperative evacuation is conducted on multiple risk virtual machines belonging to a single virtual network, so as to shorten the evacuation time of that virtual network and reduce its damage risk. Simulation results show that the proposed mechanism can achieve better performance than its counterparts in terms of virtual network evacuation completion ratio and average evacuation completion time in different observation periods.
2021, 43(10): 2894-2901.
doi: 10.11999/JEIT200711
Abstract:
Novel signal detection technique is conceived for Generalized Space Shift Keying (GSSK) modulated indoor Visible Light Communication (VLC) system, which is aided by one of popular machine learning approach termed as Support Vector Machine (SVM). For general classic VLC system, as the optimal detection algorithm, Maximum Likelihood (ML) detection has a high computational complexity. In order to alleviate this problem, classification idea in SVM is utilized to realize signal detection at the user’s receiving end by a particular trained learning model. As a result, a signal detection algorithm for the considered GSSK-VLC system based on SVM is designed with lower computational complexity and nearly optimal detection accuracy. Simulation results demonstrate that the proposed SVM-aided signal detection technique can have near optima ML Bit Error Rate (BER) performance while the computational complexity is significantly reduced in the considered indoor GSSK-VLC system
Novel signal detection technique is conceived for Generalized Space Shift Keying (GSSK) modulated indoor Visible Light Communication (VLC) system, which is aided by one of popular machine learning approach termed as Support Vector Machine (SVM). For general classic VLC system, as the optimal detection algorithm, Maximum Likelihood (ML) detection has a high computational complexity. In order to alleviate this problem, classification idea in SVM is utilized to realize signal detection at the user’s receiving end by a particular trained learning model. As a result, a signal detection algorithm for the considered GSSK-VLC system based on SVM is designed with lower computational complexity and nearly optimal detection accuracy. Simulation results demonstrate that the proposed SVM-aided signal detection technique can have near optima ML Bit Error Rate (BER) performance while the computational complexity is significantly reduced in the considered indoor GSSK-VLC system
2021, 43(10): 2902-2910.
doi: 10.11999/JEIT200995
Abstract:
In this paper, a match theory based uplink user association algorithm is proposed for maximizing energy efficiency and guaranteeing Quality of Service (QoS) requirements in a ultra-dense heterogeneous cellular networks with Human-to-Human (H2H) and Machine-to-Machine (M2M) coexistence. To maximize energy efficiency, balance load and guarantee QoS, simultaneously, the algorithm considers two access mechanisms for different types of node. Simulation results show that the proposed algorithm not only outperforms existing schemes in terms of the overall energy efficiency, the load balance and QoS guarantees, but also achieves the same performance as exhaustive search. Furthermore, the convergence speed of the algorithm is very fast, and is not affected by the number of base stations and nodes. Thus, the algorithm is suitable to solve the problem of user association, while considering the networks with H2H and M2M coexistence.
In this paper, a match theory based uplink user association algorithm is proposed for maximizing energy efficiency and guaranteeing Quality of Service (QoS) requirements in a ultra-dense heterogeneous cellular networks with Human-to-Human (H2H) and Machine-to-Machine (M2M) coexistence. To maximize energy efficiency, balance load and guarantee QoS, simultaneously, the algorithm considers two access mechanisms for different types of node. Simulation results show that the proposed algorithm not only outperforms existing schemes in terms of the overall energy efficiency, the load balance and QoS guarantees, but also achieves the same performance as exhaustive search. Furthermore, the convergence speed of the algorithm is very fast, and is not affected by the number of base stations and nodes. Thus, the algorithm is suitable to solve the problem of user association, while considering the networks with H2H and M2M coexistence.
2021, 43(10): 2911-2919.
doi: 10.11999/JEIT201005
Abstract:
The traditional spectrum sensing method of Convolutional Neural Network (CNN) has a simple network structure which limits the ability of feature extraction. To solve the problem of gradient disappearance, a cooperative spectrum sensing method based on Deep Convolutional Neural Network (DCNN) is proposed in this paper, in which shortcut connections are added to the CNN to realize the deeper network of input level identity radiation. This method transforms the spectrum sensing problem into the image binary classification problem, and performs normalized gray level processing on the covariance matrix of Quadrature Phase Shift Keying (QPSK) signal as the input of DCNN, which trains DCNN model through residual learning and extracts the deep image features of the two-dimensional grayscale image. The testing data is input into the trained model and spectrum sensing based on image classification is completed. The experimental results show that the proposed method has higher detection probability and lower false alarm probability than the traditional spectrum sensing method when the Signal to Noise Ratio (SNR) is low and multiple users collaborate in sensing.
The traditional spectrum sensing method of Convolutional Neural Network (CNN) has a simple network structure which limits the ability of feature extraction. To solve the problem of gradient disappearance, a cooperative spectrum sensing method based on Deep Convolutional Neural Network (DCNN) is proposed in this paper, in which shortcut connections are added to the CNN to realize the deeper network of input level identity radiation. This method transforms the spectrum sensing problem into the image binary classification problem, and performs normalized gray level processing on the covariance matrix of Quadrature Phase Shift Keying (QPSK) signal as the input of DCNN, which trains DCNN model through residual learning and extracts the deep image features of the two-dimensional grayscale image. The testing data is input into the trained model and spectrum sensing based on image classification is completed. The experimental results show that the proposed method has higher detection probability and lower false alarm probability than the traditional spectrum sensing method when the Signal to Noise Ratio (SNR) is low and multiple users collaborate in sensing.
2021, 43(10): 2920-2927.
doi: 10.11999/JEIT200702
Abstract:
Considering the problems of multihop relay transmission in energy harvesting cognitive radio networks, a novel Power Beacon (PB) assisted energy harvesting cognitive multihop relay network model with primary network interference is proposed, and a one-way transmission scheme is proposed. In the scenario of interference link statistical channel state information, the closed-form formulas of exact and asymptotic total outage probability are derived. In view of the complexity and nonconvexity of the exact total outage probability expression, the Adaptive Chaos Particle Swarm Optimization (ACPSO) algorithm is used to optimize the total outage performance of the secondary network. Simulation results show that the parameters such as PB’s power, interference constraint, number of secondary network hops, energy harvesting ratio, the number of primary receivers and channel capacity threshold have significant impacts on outage performance, the proposed algorithm can quickly and effectively optimize the network outage performance.
Considering the problems of multihop relay transmission in energy harvesting cognitive radio networks, a novel Power Beacon (PB) assisted energy harvesting cognitive multihop relay network model with primary network interference is proposed, and a one-way transmission scheme is proposed. In the scenario of interference link statistical channel state information, the closed-form formulas of exact and asymptotic total outage probability are derived. In view of the complexity and nonconvexity of the exact total outage probability expression, the Adaptive Chaos Particle Swarm Optimization (ACPSO) algorithm is used to optimize the total outage performance of the secondary network. Simulation results show that the parameters such as PB’s power, interference constraint, number of secondary network hops, energy harvesting ratio, the number of primary receivers and channel capacity threshold have significant impacts on outage performance, the proposed algorithm can quickly and effectively optimize the network outage performance.
2021, 43(10): 2928-2935.
doi: 10.11999/JEIT200682
Abstract:
For the problem in the existing Multi-Layer Perceptron (MLP) based DDoS detection model parameter update method that the old model parameter training dataset knowledge is forgettable and the time and space complexity are enormous, a novel model parameter UpDate method EWC-UD based on Elastic Weight Consolidation (EWC) is proposed. Firstly, the cluster center points of the old dataset are calculated as the calculation samples of Fisher information matrix by the K-Means algorithm. The coverage rates of clusters and sampling uniformity are raised effectively, which significantly reduces the amount of Fisher Information Matrix calculation and improves the efficiency of the model parameter updates. Secondly, according to the calculated Fisher information matrix, a secondary penalty item is added to the loss function, limiting the important weight and bias parameter changes in the neural network. Maintaining the detection performance of the old DDoS attack dataset, EWC-UD improves the detection accuracy of the new DDoS attack datasets. Then based on probability theory, the correctness of EWC-UD is proved, and the time complexity is analyzed. Experiments show that for the constructed test dataset, the detection accuracy of EWC-UD is 37.05% higher than the MLP-UD that only trains the new DDoS attack dataset, and compared with the time MLP-UD training both new and old DDoS attack datasets, the time and memory costs are reduced by 80.65% and 33.18 respectively.
For the problem in the existing Multi-Layer Perceptron (MLP) based DDoS detection model parameter update method that the old model parameter training dataset knowledge is forgettable and the time and space complexity are enormous, a novel model parameter UpDate method EWC-UD based on Elastic Weight Consolidation (EWC) is proposed. Firstly, the cluster center points of the old dataset are calculated as the calculation samples of Fisher information matrix by the K-Means algorithm. The coverage rates of clusters and sampling uniformity are raised effectively, which significantly reduces the amount of Fisher Information Matrix calculation and improves the efficiency of the model parameter updates. Secondly, according to the calculated Fisher information matrix, a secondary penalty item is added to the loss function, limiting the important weight and bias parameter changes in the neural network. Maintaining the detection performance of the old DDoS attack dataset, EWC-UD improves the detection accuracy of the new DDoS attack datasets. Then based on probability theory, the correctness of EWC-UD is proved, and the time complexity is analyzed. Experiments show that for the constructed test dataset, the detection accuracy of EWC-UD is 37.05% higher than the MLP-UD that only trains the new DDoS attack dataset, and compared with the time MLP-UD training both new and old DDoS attack datasets, the time and memory costs are reduced by 80.65% and 33.18 respectively.
2021, 43(10): 2936-2943.
doi: 10.11999/JEIT210095
Abstract:
Based on the theory of generalized cyclotomy, the minimal polynomial and linear complexity of a class of generalized cyclotomic quaternary sequences with period\begin{document}$2{p^2}$\end{document} ![]()
![]()
\begin{document}${F_q}$\end{document} ![]()
![]()
\begin{document}$q = {r^m}$\end{document} ![]()
![]()
\begin{document}${p^2}$\end{document} ![]()
![]()
\begin{document}$2{p^2}$\end{document} ![]()
![]()
Based on the theory of generalized cyclotomy, the minimal polynomial and linear complexity of a class of generalized cyclotomic quaternary sequences with period
2021, 43(10): 2944-2951.
doi: 10.11999/JEIT200679
Abstract:
To improve the accuracy of malicious domain name detection, a new detection model based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) is proposed. The model extracts the sequence features from different length strings to classify the domain name. Firstly, in view of the sparseness of the N-Gram feature, the model utilizes CNN with different kernels to preserve the local association between the characters in the domain name strings and convert it to dense feature vectors. Secondly, in order to mine the context information of the domain name strings, LSTM is used to extract the deep-level sequence features of different character combinations. A sequence feature attention module is designed to assign little weight value to the sequence feature extracted from the padding characters, which decreases the interference by the padding characters and enhances the ability to capture distant sequence features. Finally, combining the advantages of CNN to extract local features and LSTM to extract sequence features, both partial and sequential information are put forward to improving the detection performance. Experimental results show that the recall rate and the F1-score of the proposed model are superior to other comparative models which are solely composed of CNN or LSTM. Particularly, when dealing with the matsnu and suppobox, the proposed model has increased by 24.8% and 3.77% in accuracy compared with the model based on LSTM, respectively.
To improve the accuracy of malicious domain name detection, a new detection model based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) is proposed. The model extracts the sequence features from different length strings to classify the domain name. Firstly, in view of the sparseness of the N-Gram feature, the model utilizes CNN with different kernels to preserve the local association between the characters in the domain name strings and convert it to dense feature vectors. Secondly, in order to mine the context information of the domain name strings, LSTM is used to extract the deep-level sequence features of different character combinations. A sequence feature attention module is designed to assign little weight value to the sequence feature extracted from the padding characters, which decreases the interference by the padding characters and enhances the ability to capture distant sequence features. Finally, combining the advantages of CNN to extract local features and LSTM to extract sequence features, both partial and sequential information are put forward to improving the detection performance. Experimental results show that the recall rate and the F1-score of the proposed model are superior to other comparative models which are solely composed of CNN or LSTM. Particularly, when dealing with the matsnu and suppobox, the proposed model has increased by 24.8% and 3.77% in accuracy compared with the model based on LSTM, respectively.
2021, 43(10): 2952-2957.
doi: 10.11999/JEIT200764
Abstract:
A DNA computing algorithm is proposed in this paper which uses the assembly process of DNA/Au nanoparticle conjugation to solve an NP-complete problem in the Graph theory, the Maximum Matching Problem. According to the algorithm, the special DNA/Au nanoparticle conjugation is designed, which assembled based on a specific graph. Then, a series of experimental techniques are utilized to get the final result. This biochemical algorithm can reduce the complexity of the maximum matching problem greatly, which will provide a practical way to the best use of DNA self-assembly model.
A DNA computing algorithm is proposed in this paper which uses the assembly process of DNA/Au nanoparticle conjugation to solve an NP-complete problem in the Graph theory, the Maximum Matching Problem. According to the algorithm, the special DNA/Au nanoparticle conjugation is designed, which assembled based on a specific graph. Then, a series of experimental techniques are utilized to get the final result. This biochemical algorithm can reduce the complexity of the maximum matching problem greatly, which will provide a practical way to the best use of DNA self-assembly model.
2021, 43(10): 2958-2966.
doi: 10.11999/JEIT200694
Abstract:
For the problem of insufficient accuracy of stereo image quality prediction, a blind stereoscopic image quality assessment model combining spatial domain and transform domain to extract quality-aware features is proposed. Firstly, the statistical features of the natural scenes in the left and right views are extracted respectively in space domain and transformation domain, and statistical features of natural scenes from synthetic monocular images is extracted in transformation domain. Finally, Support Vector Regression (SVR) is used to train a stereoscopic image quality evaluation model from the feature domain to the quality score domain, so as to establish SIQA objective quality evaluation model. The performance of the proposed method is compared with some state-of-the-art full-reference, reduced-reference and no-reference stereoscopic image quality evaluation algorithms on the four public stereo image databases, taking the performance test in live 3D phase I image library as an example. SROCC of 0.967, PLCC of 0.946 and RMSE of 5.603 are achieved, which verifies the effectiveness of the proposed algorithm.
For the problem of insufficient accuracy of stereo image quality prediction, a blind stereoscopic image quality assessment model combining spatial domain and transform domain to extract quality-aware features is proposed. Firstly, the statistical features of the natural scenes in the left and right views are extracted respectively in space domain and transformation domain, and statistical features of natural scenes from synthetic monocular images is extracted in transformation domain. Finally, Support Vector Regression (SVR) is used to train a stereoscopic image quality evaluation model from the feature domain to the quality score domain, so as to establish SIQA objective quality evaluation model. The performance of the proposed method is compared with some state-of-the-art full-reference, reduced-reference and no-reference stereoscopic image quality evaluation algorithms on the four public stereo image databases, taking the performance test in live 3D phase I image library as an example. SROCC of 0.967, PLCC of 0.946 and RMSE of 5.603 are achieved, which verifies the effectiveness of the proposed algorithm.
2021, 43(10): 2967-2975.
doi: 10.11999/JEIT200531
Abstract:
The forensics methods of DeepFake video generally use convolution neural networks. However, these methods perform poorly on hard compressed DeepFake datasets and make a large number of false detections on real data. To solve the problem above, a method of hard compressed DeepFake video detection based on deep neural network model is proposed, which improves the detection accuracy of hard compressed video by incorporating super-resolution reconstruction technology and recovering the loss of the spatial and temporal information during hard compression. Experiments are performed with the FaceForensics++ Datasets and DFDC (the DeepFake Detection Challenge) Datasets for hard compressed DeepFake video, which improve the test accuracy of single frame and video compared to ResNet50, and effectively alleviate the problem of false detection of real video with hard compression.
The forensics methods of DeepFake video generally use convolution neural networks. However, these methods perform poorly on hard compressed DeepFake datasets and make a large number of false detections on real data. To solve the problem above, a method of hard compressed DeepFake video detection based on deep neural network model is proposed, which improves the detection accuracy of hard compressed video by incorporating super-resolution reconstruction technology and recovering the loss of the spatial and temporal information during hard compression. Experiments are performed with the FaceForensics++ Datasets and DFDC (the DeepFake Detection Challenge) Datasets for hard compressed DeepFake video, which improve the test accuracy of single frame and video compared to ResNet50, and effectively alleviate the problem of false detection of real video with hard compression.
2021, 43(10): 2976-2984.
doi: 10.11999/JEIT200590
Abstract:
In view of the problems of low quality, blurred borders and excessive artifacts generated by unsupervised monocular depth estimation, a deep network encoder-decoder structure based on dense feature fusion is proposed. A Dense Feature Fusion Layer(DFFL) is designed and it is filled with U-shaped encoder-decoder in the form of dense connection, while simplifying the encoder part to achieve a balanced performance of the encoder and decoder. During the training process, the calibrated stereo pair is input to the network to constrain the network to generate disparity maps by the similarity of reconstructed views. During the test process, the generated disparity map is converted into a depth map based on the known camera baseline distance and focal length. The experimental results on the KITTI data set show that this method is superior to the existing algorithms in terms of prediction accuracy and error value.
In view of the problems of low quality, blurred borders and excessive artifacts generated by unsupervised monocular depth estimation, a deep network encoder-decoder structure based on dense feature fusion is proposed. A Dense Feature Fusion Layer(DFFL) is designed and it is filled with U-shaped encoder-decoder in the form of dense connection, while simplifying the encoder part to achieve a balanced performance of the encoder and decoder. During the training process, the calibrated stereo pair is input to the network to constrain the network to generate disparity maps by the similarity of reconstructed views. During the test process, the generated disparity map is converted into a depth map based on the known camera baseline distance and focal length. The experimental results on the KITTI data set show that this method is superior to the existing algorithms in terms of prediction accuracy and error value.
2021, 43(10): 2985-2992.
doi: 10.11999/JEIT200741
Abstract:
A new object contour partition model based on the fully convolutional network, combined with the idea of generative counter network and consistent attributes is proposed. Firstly, the image region partition network is used as a generator to divide the image region. Then the structural similarity is used as the reconstruction loss of regional division to supervise and guide model learning from the perspective of visual system. Finally, the global and local context discrimination networks are used as double-path similarity to supervise the reconstruction loss of regional division and guide model learning from the discriminators to distinguish the truth and falsity of the results of regional division, and a joint loss is proposed to train the supervision model in combination with the adversarial loss, so as to make the content of regional division true, natural and with attribute consistency. The instantaneity and effectiveness of the method are verified by living examples.
A new object contour partition model based on the fully convolutional network, combined with the idea of generative counter network and consistent attributes is proposed. Firstly, the image region partition network is used as a generator to divide the image region. Then the structural similarity is used as the reconstruction loss of regional division to supervise and guide model learning from the perspective of visual system. Finally, the global and local context discrimination networks are used as double-path similarity to supervise the reconstruction loss of regional division and guide model learning from the discriminators to distinguish the truth and falsity of the results of regional division, and a joint loss is proposed to train the supervision model in combination with the adversarial loss, so as to make the content of regional division true, natural and with attribute consistency. The instantaneity and effectiveness of the method are verified by living examples.
2021, 43(10): 2993-3000.
doi: 10.11999/JEIT200964
Abstract:
This paper proposes a Behavior Delayed Sharing Network (BDSN) model to solve the personalized product recommendation problem based on personal historical browsing behaviors. First, a Behavior Delay Gated Recurrent Neural Unit (BDGRU) is presented, which uses the historical browsing time interval as a user activity factor, and updates the neuron state to calculate the user's interest expression. Then, a shared parameter network is proposed to converge the representation vectors on the user side and the goods side into a unified space. Experiments show that the AUC index and loss function of BDSN model on the validation set are both optimal, and the AUC index on the test set increases by 37% compared with the basic model.
This paper proposes a Behavior Delayed Sharing Network (BDSN) model to solve the personalized product recommendation problem based on personal historical browsing behaviors. First, a Behavior Delay Gated Recurrent Neural Unit (BDGRU) is presented, which uses the historical browsing time interval as a user activity factor, and updates the neuron state to calculate the user's interest expression. Then, a shared parameter network is proposed to converge the representation vectors on the user side and the goods side into a unified space. Experiments show that the AUC index and loss function of BDSN model on the validation set are both optimal, and the AUC index on the test set increases by 37% compared with the basic model.
2021, 43(10): 3001-3009.
doi: 10.11999/JEIT200821
Abstract:
Optimal granularity selection is one of the hotspots in the research of sequential three-way decisions. It aims to solve complex problems through reasonable granularity selection. At present, in the field of optimal granularity selection, cost sensitivity is one of the important factors affecting decision making. To solve this problem, firstly, based on information gain and chi-squared test, a novel method to measure the attribute significance is proposed when constructing the multi-granularity space in this paper. Then, to better conform the practical application, the corresponding penalty rule is set by combining the cost parameters and the granularity, and the variation rule of the decision threshold is analyzed. Finally, to eliminate the influence of the dimensional difference between the test cost and the decision cost, an objective cost calculation method is given by the coefficient of variation. The experimental results show that the proposed algorithm can be used in existing cost cognition scene, and the optimal granular layer with the lowest cost can be obtained under the given cost scene.
Optimal granularity selection is one of the hotspots in the research of sequential three-way decisions. It aims to solve complex problems through reasonable granularity selection. At present, in the field of optimal granularity selection, cost sensitivity is one of the important factors affecting decision making. To solve this problem, firstly, based on information gain and chi-squared test, a novel method to measure the attribute significance is proposed when constructing the multi-granularity space in this paper. Then, to better conform the practical application, the corresponding penalty rule is set by combining the cost parameters and the granularity, and the variation rule of the decision threshold is analyzed. Finally, to eliminate the influence of the dimensional difference between the test cost and the decision cost, an objective cost calculation method is given by the coefficient of variation. The experimental results show that the proposed algorithm can be used in existing cost cognition scene, and the optimal granular layer with the lowest cost can be obtained under the given cost scene.
2021, 43(10): 3010-3017.
doi: 10.11999/JEIT200904
Abstract:
Because single mode ElectroCardioGraph (ECG) and PhotoPlethysmoGraph(PPG) existed problem with the low recognition accuracy, not considering intra-class correlation, this paper proposes a recognition method based on the Discriminant Correlation Analysis (DCA) for the feature layer fusion of the ECG and PPG combined feature matrix and the fusion of the K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) classifiers at the decision layer. The experimental results show that the use of fusion features (ECG-PPG) and fusion the classifier (KNN-SVM) method can classify and recognize 23 subjects with an accuracy of 98.2%, and the recognition accuracy is better than single-modal recognition in the conventional environment. It provides an effective model for multimodal biometric identification.
Because single mode ElectroCardioGraph (ECG) and PhotoPlethysmoGraph(PPG) existed problem with the low recognition accuracy, not considering intra-class correlation, this paper proposes a recognition method based on the Discriminant Correlation Analysis (DCA) for the feature layer fusion of the ECG and PPG combined feature matrix and the fusion of the K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) classifiers at the decision layer. The experimental results show that the use of fusion features (ECG-PPG) and fusion the classifier (KNN-SVM) method can classify and recognize 23 subjects with an accuracy of 98.2%, and the recognition accuracy is better than single-modal recognition in the conventional environment. It provides an effective model for multimodal biometric identification.
2021, 43(10): 3018-3027.
doi: 10.11999/JEIT200774
Abstract:
The accurate classification of ElectroCardioGram (ECG) signals is essential for the automatic diagnosis of heart disease. In order to realize the intelligent classification of arrhythmia, an accurate classification method based on wavelet decomposition and 1D-GoogLeNet is proposed. In this method, Db6 wavelet is used to decompose the ECG signal in eight levels to obtain multi-dimensional data containing both time domain information and frequency domain information. Subsequently, Decomposed samples are used as input to 1D-GoogLeNet to train the model. In the proposed 1D-GoogLeNet model, using Inception's excellent performance in image feature extraction, the two-dimensional convolution is transformed into one-dimensional convolution to learn the features of ECG, and the structure of each Inception is simplified, and the model parameters are reduced. The deep learning classifier proposed in this paper can effectively alleviate the problems of low computational efficiency, difficulty in convergence and model degradation. In the experiment, the MIT-BIH arrhythmia dataset is used to test the performance of the proposed model. The experiment compares the detection results when different decomposition component combinations are used as input. When the input data is combined by {d2-d7}, the proposed 1D-GoogLeNet model can achieve an average accuracy of 96.58%. In addition, the performance of the model and the simple one-dimensional GoogLeNet without structural optimization on the data set is compared. The accuracy of the former is 4.7% higher than the latter, and the training efficiency is increased by 118%.
The accurate classification of ElectroCardioGram (ECG) signals is essential for the automatic diagnosis of heart disease. In order to realize the intelligent classification of arrhythmia, an accurate classification method based on wavelet decomposition and 1D-GoogLeNet is proposed. In this method, Db6 wavelet is used to decompose the ECG signal in eight levels to obtain multi-dimensional data containing both time domain information and frequency domain information. Subsequently, Decomposed samples are used as input to 1D-GoogLeNet to train the model. In the proposed 1D-GoogLeNet model, using Inception's excellent performance in image feature extraction, the two-dimensional convolution is transformed into one-dimensional convolution to learn the features of ECG, and the structure of each Inception is simplified, and the model parameters are reduced. The deep learning classifier proposed in this paper can effectively alleviate the problems of low computational efficiency, difficulty in convergence and model degradation. In the experiment, the MIT-BIH arrhythmia dataset is used to test the performance of the proposed model. The experiment compares the detection results when different decomposition component combinations are used as input. When the input data is combined by {d2-d7}, the proposed 1D-GoogLeNet model can achieve an average accuracy of 96.58%. In addition, the performance of the model and the simple one-dimensional GoogLeNet without structural optimization on the data set is compared. The accuracy of the former is 4.7% higher than the latter, and the training efficiency is increased by 118%.
2021, 43(10): 3028-3034.
doi: 10.11999/JEIT200615
Abstract:
Feature selection is an essential step in the data preprocessing phase in the fields of machine learning, natural language processing and data mining. In some feature selection algorithms based on information theory, there is a problem that choosing different parameters means choosing different feature selection algorithms. How to determine the dynamic, non-a priori weights and avoid the preset a priori parameters become an urgent problem. A Dynamic Weighted Maximum Relevance and maximum Independence (WMRI) feature selection algorithm is proposed in this paper. Firstly, the algorithm calculates the average value of the new classification information and the retained classification information. Secondly, the standard deviation is used to dynamically adjust the parameter weights of these two types of classification information. At last, WMRI and the other five feature selection algorithms use ten different data sets on three classifiers for the fmi classification metrics validation. The experimental results show that the WMRI method can improve the quality of feature subsets and increase classification accuracy.
Feature selection is an essential step in the data preprocessing phase in the fields of machine learning, natural language processing and data mining. In some feature selection algorithms based on information theory, there is a problem that choosing different parameters means choosing different feature selection algorithms. How to determine the dynamic, non-a priori weights and avoid the preset a priori parameters become an urgent problem. A Dynamic Weighted Maximum Relevance and maximum Independence (WMRI) feature selection algorithm is proposed in this paper. Firstly, the algorithm calculates the average value of the new classification information and the retained classification information. Secondly, the standard deviation is used to dynamically adjust the parameter weights of these two types of classification information. At last, WMRI and the other five feature selection algorithms use ten different data sets on three classifiers for the fmi classification metrics validation. The experimental results show that the WMRI method can improve the quality of feature subsets and increase classification accuracy.
2021, 43(10): 3035-3042.
doi: 10.11999/JEIT200510
Abstract:
With the current development of mobile terminal devices and the popularity of 5G technology, there is an increasing demand for mobile group intelligence awareness. However, the current distribution methods for sensing tasks still suffer from inefficient, costly and unstable transmission, which limits greatly the completion of sensing terminal tasks. For this reason, an Incentive Cost Task Distribution Model (ICTDM) and a User Number Task Distribution Model (UNTDM) based on the good coverage of urban rail transit and the predictability of urban rail transit are proposed. The selective distribution of sensory tasks in different areas of the city is achieved through the sparseness of rail traffic for aggregated pedestrian flows. And the minimization of the number of people required for the task and the distance moved is used as a means to accomplish the purpose of reducing the total incentive cost of the system. Experimental results show that this algorithm can achieve fewer task participant distribution schemes by optimizing the task distribution process to reduce the cost of perceived tasks compared with similar algorithms, while completing the same set of tasks.
With the current development of mobile terminal devices and the popularity of 5G technology, there is an increasing demand for mobile group intelligence awareness. However, the current distribution methods for sensing tasks still suffer from inefficient, costly and unstable transmission, which limits greatly the completion of sensing terminal tasks. For this reason, an Incentive Cost Task Distribution Model (ICTDM) and a User Number Task Distribution Model (UNTDM) based on the good coverage of urban rail transit and the predictability of urban rail transit are proposed. The selective distribution of sensory tasks in different areas of the city is achieved through the sparseness of rail traffic for aggregated pedestrian flows. And the minimization of the number of people required for the task and the distance moved is used as a means to accomplish the purpose of reducing the total incentive cost of the system. Experimental results show that this algorithm can achieve fewer task participant distribution schemes by optimizing the task distribution process to reduce the cost of perceived tasks compared with similar algorithms, while completing the same set of tasks.
2021, 43(10): 3043-3049.
doi: 10.11999/JEIT200623
Abstract:
In order to quickly and accurately suppress the howling effect in hearing aids, this paper presents an analytical design of all-phase Finite Impulse Response(FIR) notch filter with explicit controllable center frequency. Firstly, to obtain the higher accuracy, integer m and decimal λ are introduced to control the central frequency of the notch filter. Then, an even symmetric closed-form analytic formula is designed to complete the design of the notch filter, which shows that the proposed notch filter has linear transmission characteristics and avoids nonlinear distortion. Finally, data extension and interception are carried out to ensure the continuity and linear phase of the output signal. Herein, the proposed notch filter is inserted into hearing aid to suppress the howling for the sake of verifying its performances. The experimental results show that the attenuation value of the proposed filter at the howling frequency can reach –330 dB, and the SNR is 22 dB. Moreover, the proposed filter is of good output waveform quality, low algorithm complexity and high robustness, and it has a certain application prospect.
In order to quickly and accurately suppress the howling effect in hearing aids, this paper presents an analytical design of all-phase Finite Impulse Response(FIR) notch filter with explicit controllable center frequency. Firstly, to obtain the higher accuracy, integer m and decimal λ are introduced to control the central frequency of the notch filter. Then, an even symmetric closed-form analytic formula is designed to complete the design of the notch filter, which shows that the proposed notch filter has linear transmission characteristics and avoids nonlinear distortion. Finally, data extension and interception are carried out to ensure the continuity and linear phase of the output signal. Herein, the proposed notch filter is inserted into hearing aid to suppress the howling for the sake of verifying its performances. The experimental results show that the attenuation value of the proposed filter at the howling frequency can reach –330 dB, and the SNR is 22 dB. Moreover, the proposed filter is of good output waveform quality, low algorithm complexity and high robustness, and it has a certain application prospect.
2021, 43(10): 3050-3057.
doi: 10.11999/JEIT200874
Abstract:
A hybrid multi-level hardware Trojan detection method based on XGBoost algorithm is proposed for the problem of hardware Trojans implanted by malicious third-party manufacturers. The detection method treats each wire in gate-level netlist as a node and detects Trojans in three levels. Firstly, the effective static features of the circuit are extracted and the XGBoost algorithm is applied to detect the suspicious Trojan circuits. Common circuits distinguished at the first level continued to be detected at the second level by analyzing scan chain structural features. Finally, dynamic detection is used to increase further the accuracy of Trojans detection. Experimental results on Trust-hub benchmark show that this method has a higher accuracy compared with other existing detection methods. This detection method can finally achieve 94.0% average True Positive Rate (TPR) and 99.3% average True Negative Rate (TNR).
A hybrid multi-level hardware Trojan detection method based on XGBoost algorithm is proposed for the problem of hardware Trojans implanted by malicious third-party manufacturers. The detection method treats each wire in gate-level netlist as a node and detects Trojans in three levels. Firstly, the effective static features of the circuit are extracted and the XGBoost algorithm is applied to detect the suspicious Trojan circuits. Common circuits distinguished at the first level continued to be detected at the second level by analyzing scan chain structural features. Finally, dynamic detection is used to increase further the accuracy of Trojans detection. Experimental results on Trust-hub benchmark show that this method has a higher accuracy compared with other existing detection methods. This detection method can finally achieve 94.0% average True Positive Rate (TPR) and 99.3% average True Negative Rate (TNR).
2021, 43(10): 3058-3067.
doi: 10.11999/JEIT210087
Abstract:
Developing new active substance composition system and its preparation method to enhance scandate cathode’s emission property is a hotspot in the research field of thermionic cathode especially high-emission cathode. A novel highly active impregnant substance consisting of polymetallic oxide which apparently increases scandium’s appending proportion and greatly enhances cathode’s emission current density is put forward in this paper. Freeze-drying method is applied into preparation of the active substance’s precursor and effectively solves the problem of inhomogeneity and uncontrollability in the mechanical crushing, grinding and mixing procedures of conventional solid-phase synthesis routine. Cathode which adopted novel composition system and substance acquired by new preparation routine reaches a pulse emission current density of above 500 A/cm2 under close-spaced diode configuration and 218.5 A/cm2 in an electron gun. Under the DC diode experimental configuration, the cathodes’ emission lifetime test has endured for 10500 hours with no emission current drop; while in the electron gun with a pulse drive of heavy duty cycle (5%), the cathode maintains a big workload of more than 50 A/cm2 after having worked for 2010 hours. Via Deep UltraViolet laser-Photo Emission and Thermal Emission Electron Microscopy (DUV-PEEM/TEEM) analyzation, the phenomenon ofthermionic emission points’ amount increases and emitting micro-area expands on the newly prepared high-emission cathode’s surface is observed. Finally, a ‘binary tree’ emission model is brought up, hoping to explain the physical mechanism of scandate cathode’s high emission character with new active substance.
Developing new active substance composition system and its preparation method to enhance scandate cathode’s emission property is a hotspot in the research field of thermionic cathode especially high-emission cathode. A novel highly active impregnant substance consisting of polymetallic oxide which apparently increases scandium’s appending proportion and greatly enhances cathode’s emission current density is put forward in this paper. Freeze-drying method is applied into preparation of the active substance’s precursor and effectively solves the problem of inhomogeneity and uncontrollability in the mechanical crushing, grinding and mixing procedures of conventional solid-phase synthesis routine. Cathode which adopted novel composition system and substance acquired by new preparation routine reaches a pulse emission current density of above 500 A/cm2 under close-spaced diode configuration and 218.5 A/cm2 in an electron gun. Under the DC diode experimental configuration, the cathodes’ emission lifetime test has endured for 10500 hours with no emission current drop; while in the electron gun with a pulse drive of heavy duty cycle (5%), the cathode maintains a big workload of more than 50 A/cm2 after having worked for 2010 hours. Via Deep UltraViolet laser-Photo Emission and Thermal Emission Electron Microscopy (DUV-PEEM/TEEM) analyzation, the phenomenon ofthermionic emission points’ amount increases and emitting micro-area expands on the newly prepared high-emission cathode’s surface is observed. Finally, a ‘binary tree’ emission model is brought up, hoping to explain the physical mechanism of scandate cathode’s high emission character with new active substance.
2021, 43(10): 3068-3074.
doi: 10.11999/JEIT200571
Abstract:
Electrostatic field measurement is the most direct method to detect whether an object carries excessive static charge. Traditional electrostatic meters mainly rely on fixed distance measurement. While the measured object is difficult to maintain still or not easy to approach, the distance change causes a sensitivity change and results in a measurement error. Based on the micro-machined electric field sensor, this paper proposes an electrostatic measurement idea that the meter adjusts the sensitivity adaptively according to the tested distance: measure the distance of the object through the ultrasonic module, to find the corresponding sensitivity coefficient through the microprocessor, and then to calculate the measured voltage accompany the electric field result. For the developed meter, this paper proposes a calibration method based on a combination of laboratory calibration and site calibration, builds a dynamic sensitivity calibration system, and calculates the corresponding relation of sensor sensitivity coefficients of different test distances and different measured object sizes. Compared with the conventional measurements in a fixed distance, this paper achieves more accurate non-contact surface static voltage measurement by means of sensitivity dynamic calibration. In the meanwhile, the micro electric field sensing element has the advantages of small size, low power consumption, easy integration, and capable of mass manufacturing. According to the third-party test, the average error at different test distances is –2.98%.
Electrostatic field measurement is the most direct method to detect whether an object carries excessive static charge. Traditional electrostatic meters mainly rely on fixed distance measurement. While the measured object is difficult to maintain still or not easy to approach, the distance change causes a sensitivity change and results in a measurement error. Based on the micro-machined electric field sensor, this paper proposes an electrostatic measurement idea that the meter adjusts the sensitivity adaptively according to the tested distance: measure the distance of the object through the ultrasonic module, to find the corresponding sensitivity coefficient through the microprocessor, and then to calculate the measured voltage accompany the electric field result. For the developed meter, this paper proposes a calibration method based on a combination of laboratory calibration and site calibration, builds a dynamic sensitivity calibration system, and calculates the corresponding relation of sensor sensitivity coefficients of different test distances and different measured object sizes. Compared with the conventional measurements in a fixed distance, this paper achieves more accurate non-contact surface static voltage measurement by means of sensitivity dynamic calibration. In the meanwhile, the micro electric field sensing element has the advantages of small size, low power consumption, easy integration, and capable of mass manufacturing. According to the third-party test, the average error at different test distances is –2.98%.
