Email alert
2021 Vol. 43, No. 11
Display Method:
2021, 43(11): 3075-3085.
doi: 10.11999/JEIT200899
Abstract:
It is known from electromagnetic momentum that electromagnetic waves can carry Spin Angular Momentum (SAM) related to polarization and Orbital Angular Momentum (OAM) related to the trajectory of the Poynting vector. When OAM is not zero, the wave-front electric field distribution of the electromagnetic wave is vortex-like and has the characteristic of propagating along the axial direction. Therefore, this electromagnetic wave is aptly named vortex electromagnetic wave. Based on the mathematical model of the plane electromagnetic wave field, the researchers introduce a Fourier factor that uses the topological charge (also called mode) of the OAM as a parameter to describe the field of the vortex electromagnetic wave. Therefore, the wave-front of the vortex electromagnetic wave with a “polarization” pattern associated with topological charge, the use of polarization patterns of vortex electromagnetic waves in different modes can further increase the spectrum effect of the wireless communication system. Studies show that although it is feasible to generate "planar" vortex electromagnetic wave beams from Uniform Circular Array (UCA) arrays in an open environment, to obtain modal multiplexing gain, and it is necessary to explore vortex electromagnetic wave beams based on orthogonal phase sequences distributed on a unit circle in the complex plane. At the same time, the paper also investigates the current research status of compatibility between OAM and Multiple Input Multiple Output (MIMO) systems in the field of radio frequency.
It is known from electromagnetic momentum that electromagnetic waves can carry Spin Angular Momentum (SAM) related to polarization and Orbital Angular Momentum (OAM) related to the trajectory of the Poynting vector. When OAM is not zero, the wave-front electric field distribution of the electromagnetic wave is vortex-like and has the characteristic of propagating along the axial direction. Therefore, this electromagnetic wave is aptly named vortex electromagnetic wave. Based on the mathematical model of the plane electromagnetic wave field, the researchers introduce a Fourier factor that uses the topological charge (also called mode) of the OAM as a parameter to describe the field of the vortex electromagnetic wave. Therefore, the wave-front of the vortex electromagnetic wave with a “polarization” pattern associated with topological charge, the use of polarization patterns of vortex electromagnetic waves in different modes can further increase the spectrum effect of the wireless communication system. Studies show that although it is feasible to generate "planar" vortex electromagnetic wave beams from Uniform Circular Array (UCA) arrays in an open environment, to obtain modal multiplexing gain, and it is necessary to explore vortex electromagnetic wave beams based on orthogonal phase sequences distributed on a unit circle in the complex plane. At the same time, the paper also investigates the current research status of compatibility between OAM and Multiple Input Multiple Output (MIMO) systems in the field of radio frequency.
2021, 43(11): 3086-3095.
doi: 10.11999/JEIT210115
Abstract:
Considering the intelligent decision of battlefield communication countermeasure, based on the overall confrontation, a Bootstrapped expert trajectory memory replay - Hierarchical reinforcement learning - Jamming resources distribution decision - Making algorithm(BHJM) is proposed, and the algorithm for frequency hopping jamming decision problem, according to the frequency distribution, jamming spectrum is divided, based on hierarchical reinforcement learning again decision jamming spectrum and bandwidth are divided, and finally based on the bootstrapped expert trajectory memory replay mechanism, the algorithm is optimized, the algorithm can is existing resources, especially under the condition of insufficient resources, give priority to jam the most threat target, obtain the optimal jamming effect and reduce the total jamming bandwidth. The simulation results show that, compared with the existing resource allocation decision algorithms, the proposed algorithm can save 25% of the resources of jammers and 15% of the jamming bandwidth, which is of great practical value.
Considering the intelligent decision of battlefield communication countermeasure, based on the overall confrontation, a Bootstrapped expert trajectory memory replay - Hierarchical reinforcement learning - Jamming resources distribution decision - Making algorithm(BHJM) is proposed, and the algorithm for frequency hopping jamming decision problem, according to the frequency distribution, jamming spectrum is divided, based on hierarchical reinforcement learning again decision jamming spectrum and bandwidth are divided, and finally based on the bootstrapped expert trajectory memory replay mechanism, the algorithm is optimized, the algorithm can is existing resources, especially under the condition of insufficient resources, give priority to jam the most threat target, obtain the optimal jamming effect and reduce the total jamming bandwidth. The simulation results show that, compared with the existing resource allocation decision algorithms, the proposed algorithm can save 25% of the resources of jammers and 15% of the jamming bandwidth, which is of great practical value.
2021, 43(11): 3096-3104.
doi: 10.11999/JEIT201064
Abstract:
Mobile communication applications scenarios are becoming complexity and diversity. It is difficult to have a universal transmission waveform to meet all communication needs, which puts forward high requirements for the coordination and collaboration of multiple waveforms. In this paper, an intelligent multi-carrier waveform modulation system is proposed for complex scenarios, the sending end can select a suitable transmission waveform by the waveform activation factor, the receiving end will take the I/Q component of different waveform signals as an adaptive factor, and use the main component analysis method to process the data and feed it into the Intelligent Waveform Recognition Network (IWR-Net) to complete the identification of the signal. The proposed system is integrated with deep learning network and has a more unified hardware architecture. The simulation results show that the accuracy of different send waveform recognition can be as high as 98.2% in 5G multi-scenes, and it has good generalization performance in different test environments.
Mobile communication applications scenarios are becoming complexity and diversity. It is difficult to have a universal transmission waveform to meet all communication needs, which puts forward high requirements for the coordination and collaboration of multiple waveforms. In this paper, an intelligent multi-carrier waveform modulation system is proposed for complex scenarios, the sending end can select a suitable transmission waveform by the waveform activation factor, the receiving end will take the I/Q component of different waveform signals as an adaptive factor, and use the main component analysis method to process the data and feed it into the Intelligent Waveform Recognition Network (IWR-Net) to complete the identification of the signal. The proposed system is integrated with deep learning network and has a more unified hardware architecture. The simulation results show that the accuracy of different send waveform recognition can be as high as 98.2% in 5G multi-scenes, and it has good generalization performance in different test environments.
2021, 43(11): 3105-3110.
doi: 10.11999/JEIT200784
Abstract:
Near-field source localization is a potential research topic in next-generation wireless communications. Most existing methods focus on traditional subspace based methods or on-grid sparse methods. For the problem that the accuracy of subspace class method loss array aperture and sparse representation method is restricted by mesh effect, an off-grid sparse representation localization method is proposed in this paper. First, by obtaining a high-order cumulant matrix, an angle based off-grid signal model is constructed and then the alternatively iterating optimization method is employed to estimate the angles. For range estimation, a range parameter based off-grid signal model is constructed by using the angle estimation values and is solved by alternatively iterating method. Simulation results reveal that the proposed method not only possesses high estimation accuracy, but also can realize auto-pairing of angles and ranges.
Near-field source localization is a potential research topic in next-generation wireless communications. Most existing methods focus on traditional subspace based methods or on-grid sparse methods. For the problem that the accuracy of subspace class method loss array aperture and sparse representation method is restricted by mesh effect, an off-grid sparse representation localization method is proposed in this paper. First, by obtaining a high-order cumulant matrix, an angle based off-grid signal model is constructed and then the alternatively iterating optimization method is employed to estimate the angles. For range estimation, a range parameter based off-grid signal model is constructed by using the angle estimation values and is solved by alternatively iterating method. Simulation results reveal that the proposed method not only possesses high estimation accuracy, but also can realize auto-pairing of angles and ranges.
2021, 43(11): 3111-3121.
doi: 10.11999/JEIT200196
Abstract:
For time-division duplex massive MIMO systems, two new downlink precoding schemes are proposed, namely New Maximum Ratio Combining (NMRC) and New Zero-Forcing (NZF) scheme. Through theoretical analysis, the expressions of downlink signal to interference plus noise ratio and spectral efficiency of target users and non-target users under two new precoding schemes are obtained, and compared with traditional Zero-Forcing (ZF) and Maximum Ratio Combining (MRC) Precoding schemes for downlink signal to interference plus noise ratio, spectrum efficiency and bit error rate performance. The simulation results show that the proposed NMRC and NZF precoding can achieve better performance without increasing the computational complexity compared with the traditional MRC and ZF precoding. Moreover, the new precoding schemes improve the performance of target users while improving the performance of other non-target users.
For time-division duplex massive MIMO systems, two new downlink precoding schemes are proposed, namely New Maximum Ratio Combining (NMRC) and New Zero-Forcing (NZF) scheme. Through theoretical analysis, the expressions of downlink signal to interference plus noise ratio and spectral efficiency of target users and non-target users under two new precoding schemes are obtained, and compared with traditional Zero-Forcing (ZF) and Maximum Ratio Combining (MRC) Precoding schemes for downlink signal to interference plus noise ratio, spectrum efficiency and bit error rate performance. The simulation results show that the proposed NMRC and NZF precoding can achieve better performance without increasing the computational complexity compared with the traditional MRC and ZF precoding. Moreover, the new precoding schemes improve the performance of target users while improving the performance of other non-target users.
2021, 43(11): 3122-3130.
doi: 10.11999/JEIT201009
Abstract:
To deal with the dynamic nature of 5G network resource state and the difficulty of service function chain deployment under the high-dimensional network state model, an online Service function Chain Deployment method based on Deep Q network (DeePSCD) is proposed. First, to describe the dynamic nature of network resource state, the service function chain deployment is modeled as a Markov decision process. Then, the deep Q network is used to solve the online service function chain deployment problem in the high-dimensional system resource model. This method can effectively describe the dynamic changes of network resource state. Specifically, deep Q network handles the complexity of problem and determines the optimal deployment solution of service function chain. Simulation results show that the proposed method can reduce the deployment cost of the service function chain and increase the acceptance rate while meeting the service delay constraint.
To deal with the dynamic nature of 5G network resource state and the difficulty of service function chain deployment under the high-dimensional network state model, an online Service function Chain Deployment method based on Deep Q network (DeePSCD) is proposed. First, to describe the dynamic nature of network resource state, the service function chain deployment is modeled as a Markov decision process. Then, the deep Q network is used to solve the online service function chain deployment problem in the high-dimensional system resource model. This method can effectively describe the dynamic changes of network resource state. Specifically, deep Q network handles the complexity of problem and determines the optimal deployment solution of service function chain. Simulation results show that the proposed method can reduce the deployment cost of the service function chain and increase the acceptance rate while meeting the service delay constraint.
2021, 43(11): 3131-3140.
doi: 10.11999/JEIT200739
Abstract:
An Orthogonal MultiUser Short Reference High Rate DCSK (OMU-SRHR-DCSK) communication system is proposed to solve the problem of low transmission rate and low energy efficiency of traditional MultiUser Short Reference Differential Chaos Shift Keying communication system. The system shortens reference signal length to 1/P of data signal, transmits multiple users’ information bits by adding two continuous information time slots, and transmits additionally N user’ information bit in each information time slot through Hilbert transform, which greatly improves the data transmission rate of the system. The interference between users is completely eliminated mainly by combining Hilbert transform with Walsh code, thus improving the Bit Error Rate (BER) performance. The theoretical BER formula of OMU-SRHR-DCSK system over Additive White Gaussian Noise (AWGN) and L-path Rayleigh Fading Channel (RFC) are derived and verified by experiments. The experimental simulation and theoretical derivation are consistent in the two channels, which proves the correctness of theoretical derivation. Compared with the traditional multiuser short reference system, the transmission rate of the system is greatly improved, and the BER performance of the system is obviously superior over that of the traditional multiuser short reference system under the condition of the same number of transmission bit. It is proved that the system has excellent practical value and provides a good theoretical support for its application in practice.
An Orthogonal MultiUser Short Reference High Rate DCSK (OMU-SRHR-DCSK) communication system is proposed to solve the problem of low transmission rate and low energy efficiency of traditional MultiUser Short Reference Differential Chaos Shift Keying communication system. The system shortens reference signal length to 1/P of data signal, transmits multiple users’ information bits by adding two continuous information time slots, and transmits additionally N user’ information bit in each information time slot through Hilbert transform, which greatly improves the data transmission rate of the system. The interference between users is completely eliminated mainly by combining Hilbert transform with Walsh code, thus improving the Bit Error Rate (BER) performance. The theoretical BER formula of OMU-SRHR-DCSK system over Additive White Gaussian Noise (AWGN) and L-path Rayleigh Fading Channel (RFC) are derived and verified by experiments. The experimental simulation and theoretical derivation are consistent in the two channels, which proves the correctness of theoretical derivation. Compared with the traditional multiuser short reference system, the transmission rate of the system is greatly improved, and the BER performance of the system is obviously superior over that of the traditional multiuser short reference system under the condition of the same number of transmission bit. It is proved that the system has excellent practical value and provides a good theoretical support for its application in practice.
2021, 43(11): 3141-3148.
doi: 10.11999/JEIT200831
Abstract:
The existing physical layer authentication mechanism relies on the privacy of the legitimate channel. Once the attacker can manipulate or obtain legitimate channel information, the physical layer authentication mechanism will face the threat of being compromised. To overcome the above-mentioned shortcomings, a Man-In-The-Middle (MITM) pilot attack method is proposed, which attacks the physical layer authentication mechanism by controlling the channel measurement process of the legitimate parties. Firstly, the man-in-the-middle pilot attack system is modeled, and a progressive and non-sense access strategy for MITM pilot attack is given. This strategy allows the attacker to access smoothly legitimate communication. After the attacker accesses successfully, he can launch attacks on two basic physical layer authentication mechanisms: For CSI-based comparative authentication mechanisms, denial of service attacks and counterfeit access attacks can be implemented; For the CSI-based encryption authentication mechanism, the channel information can be stolen, thereby further cracking the authentication vector. This attack method is suitable for general public pilot wireless communication systems, and requires the attacker to be able to synchronize the pilot sending process of the legitimate two parties. Simulation analysis verifies the effectiveness of multiple attack methods such as the progressive and non-sense access strategy, denial of service attack, counterfeit access attack, or cracking authentication vector.
The existing physical layer authentication mechanism relies on the privacy of the legitimate channel. Once the attacker can manipulate or obtain legitimate channel information, the physical layer authentication mechanism will face the threat of being compromised. To overcome the above-mentioned shortcomings, a Man-In-The-Middle (MITM) pilot attack method is proposed, which attacks the physical layer authentication mechanism by controlling the channel measurement process of the legitimate parties. Firstly, the man-in-the-middle pilot attack system is modeled, and a progressive and non-sense access strategy for MITM pilot attack is given. This strategy allows the attacker to access smoothly legitimate communication. After the attacker accesses successfully, he can launch attacks on two basic physical layer authentication mechanisms: For CSI-based comparative authentication mechanisms, denial of service attacks and counterfeit access attacks can be implemented; For the CSI-based encryption authentication mechanism, the channel information can be stolen, thereby further cracking the authentication vector. This attack method is suitable for general public pilot wireless communication systems, and requires the attacker to be able to synchronize the pilot sending process of the legitimate two parties. Simulation analysis verifies the effectiveness of multiple attack methods such as the progressive and non-sense access strategy, denial of service attack, counterfeit access attack, or cracking authentication vector.
2021, 43(11): 3149-3155.
doi: 10.11999/JEIT201007
Abstract:
The influence of channel construction process on the algorithm performance is not considered in the existing polar code puncturing algorithms. To solve this problem, a Puncturing algorithm of Polar Code based on Gaussian Approximation (GAPPC) is proposed. Firstly, using Gaussian approximation for channel construction of polar code and analyzing the relationship between Gaussian approximation and puncturing algorithm, the modified Gaussian approximation function is derived to reduce the output value of channel construction with introduced Gaussian correction factors. Then the ordered channel reliability set is obtained by ordering the polarization subchannels under the channel construction with the modified Gaussian approximation function. Finally, the mapping rule is determined according to the relationship of channel capacity, and the puncturing bit set and frozen bit set are selected so that the puncturing polar code is completed. Experimental results show that the frame error rate and bit error rate are significantly reduced under different code lengths and bit rates.
The influence of channel construction process on the algorithm performance is not considered in the existing polar code puncturing algorithms. To solve this problem, a Puncturing algorithm of Polar Code based on Gaussian Approximation (GAPPC) is proposed. Firstly, using Gaussian approximation for channel construction of polar code and analyzing the relationship between Gaussian approximation and puncturing algorithm, the modified Gaussian approximation function is derived to reduce the output value of channel construction with introduced Gaussian correction factors. Then the ordered channel reliability set is obtained by ordering the polarization subchannels under the channel construction with the modified Gaussian approximation function. Finally, the mapping rule is determined according to the relationship of channel capacity, and the puncturing bit set and frozen bit set are selected so that the puncturing polar code is completed. Experimental results show that the frame error rate and bit error rate are significantly reduced under different code lengths and bit rates.
2021, 43(11): 3156-3165.
doi: 10.11999/JEIT200955
Abstract:
The wireless communication technology based on Orbital Angular Momentum (OAM) can greatly improve the performance of the communication system under ideal transmission conditions. However, in the actual environment, atmospheric turbulence and aperture mismatch can cause crosstalk between OAM modes and increase the Bit Error Rate (BER). In order to reduce the BER of the optical wireless OAM multiplexing system in a complex environment, an OAM multiplexing communication system based on the Vertical Bell LAyered Space Time (VBLAST-OAM) code criterion under the scenario of atmospheric turbulence and the aperture mismatch of the transceiver is established firstly. Then, the system performance are analyzed based on the Ordered Successive Interference Cancellation (OSIC), the Markov Random Field Belief Propagation (MRF-BP) algorithm and the algorithm OAM-OSIC. Simulation results show that the algorithm proposed in this paper can reduce the BER of OAM systems effectively in complex environment and the MRF-BP has the best performance. Although OAM-OSIC is a suboptimal algorithm, it has a great advantage in the running cost.
The wireless communication technology based on Orbital Angular Momentum (OAM) can greatly improve the performance of the communication system under ideal transmission conditions. However, in the actual environment, atmospheric turbulence and aperture mismatch can cause crosstalk between OAM modes and increase the Bit Error Rate (BER). In order to reduce the BER of the optical wireless OAM multiplexing system in a complex environment, an OAM multiplexing communication system based on the Vertical Bell LAyered Space Time (VBLAST-OAM) code criterion under the scenario of atmospheric turbulence and the aperture mismatch of the transceiver is established firstly. Then, the system performance are analyzed based on the Ordered Successive Interference Cancellation (OSIC), the Markov Random Field Belief Propagation (MRF-BP) algorithm and the algorithm OAM-OSIC. Simulation results show that the algorithm proposed in this paper can reduce the BER of OAM systems effectively in complex environment and the MRF-BP has the best performance. Although OAM-OSIC is a suboptimal algorithm, it has a great advantage in the running cost.
An under-Sampling Restoration Digital Predistortion Technique Based on Landweber Iteration Algorithm
2021, 43(11): 3166-3173.
doi: 10.11999/JEIT201051
Abstract:
In order to better correct the nonlinear characteristics of Power Amplifier (PA), conventional broadband Digital PreDistortion (DPD) requires usually the feedback channel bandwidth to be 5 times of the transmitting signal bandwidth, and Analog to Digital Converter (ADC) with higher sampling rate is required accordingly, which will lead to the hardware cost and energy consumption problems of digital predistortion system. To solve this problem, an Landweber iterative algorithm-based UnderSampled Restoration (USR) Digital PreDistortion method (Landweber-USR DPD) is propsed. It is processed in an internal and external loop, which can recover the full-band output signals from the undersampled PA output signals under the requirements that the sampling rate requirement of ADC is reduced from 5 times to twice. And the restored data is closer to the real PA output signals to achieve better pre-distortion effect. In experiment, a single-device Gallium Nitride (GaN) broadband class-F PA is excited by 5 MHz-LTE signal in 1.8 GHz frequency. The feedback ADC rate is set as full sampling rate (40 Msps) and undersampling rate (10 Msps), respectively. The experimental results fully demonstrate the reliability of Landweber iterative algorithm and the effectiveness of Landweber-USR DPD technology, which can provide an idea and method for effectively reducing ADC sampling rate in the engineering implementation of digital predistortion technology in broadband communication systems.
In order to better correct the nonlinear characteristics of Power Amplifier (PA), conventional broadband Digital PreDistortion (DPD) requires usually the feedback channel bandwidth to be 5 times of the transmitting signal bandwidth, and Analog to Digital Converter (ADC) with higher sampling rate is required accordingly, which will lead to the hardware cost and energy consumption problems of digital predistortion system. To solve this problem, an Landweber iterative algorithm-based UnderSampled Restoration (USR) Digital PreDistortion method (Landweber-USR DPD) is propsed. It is processed in an internal and external loop, which can recover the full-band output signals from the undersampled PA output signals under the requirements that the sampling rate requirement of ADC is reduced from 5 times to twice. And the restored data is closer to the real PA output signals to achieve better pre-distortion effect. In experiment, a single-device Gallium Nitride (GaN) broadband class-F PA is excited by 5 MHz-LTE signal in 1.8 GHz frequency. The feedback ADC rate is set as full sampling rate (40 Msps) and undersampling rate (10 Msps), respectively. The experimental results fully demonstrate the reliability of Landweber iterative algorithm and the effectiveness of Landweber-USR DPD technology, which can provide an idea and method for effectively reducing ADC sampling rate in the engineering implementation of digital predistortion technology in broadband communication systems.
2021, 43(11): 3174-3184.
doi: 10.11999/JEIT210368
Abstract:
In the research of matching navigation and precision guidance using spaceborne SAR image as the reference image and airborne/missile SAR image as the real-time image, the traditional point feature-based matching method has too many feature points, high mismatch rate, and easy affected by problems such as noise and gray level changes. A new method for matching SAR remote sensing images from coarse to fine based on salient contour features is proposed. Based on the pre-processing of SAR images, an improved Fuzzy C-Means (FCM) clustering image segmentation is used to to extract closed contour features. Then, a normalized contour center distance descriptor is constructed for two-way matching to obtain the rough matching contours with strong robustness. Finally, the improved Local Binary Pattern (LBP) operator is employed on the rough matching contours to gain the fine matching result. The experimental results demonstrate the proposed method has the advantages of high accuracy and strong robustness in the case of image rotation, spatial variation and noise interference, and is suitable for remote sensing SAR image matching.
In the research of matching navigation and precision guidance using spaceborne SAR image as the reference image and airborne/missile SAR image as the real-time image, the traditional point feature-based matching method has too many feature points, high mismatch rate, and easy affected by problems such as noise and gray level changes. A new method for matching SAR remote sensing images from coarse to fine based on salient contour features is proposed. Based on the pre-processing of SAR images, an improved Fuzzy C-Means (FCM) clustering image segmentation is used to to extract closed contour features. Then, a normalized contour center distance descriptor is constructed for two-way matching to obtain the rough matching contours with strong robustness. Finally, the improved Local Binary Pattern (LBP) operator is employed on the rough matching contours to gain the fine matching result. The experimental results demonstrate the proposed method has the advantages of high accuracy and strong robustness in the case of image rotation, spatial variation and noise interference, and is suitable for remote sensing SAR image matching.
2021, 43(11): 3185-3192.
doi: 10.11999/JEIT201028
Abstract:
Due to the weak energy of the floating small targets, it is hard to be detected in sea surface. Relying on the energy, the traditional detectors based on statistical model inevitable loss the detection performance, regardless of the correlation between the frequency domain amplitudes. Therefore, in the paper, the correlation between the frequency domain amplitudes is considered by using the graph. Firstly, the connected density is calculated by the correlation between the frequency domain amplitudes of the echo pulses. Secondly, an adjacency matrix is generated based on the correlation. Thirdly, the adjacency matrix is converted to a Laplacian matrix. Lastly, the maximum eigenvalue of the Laplacian matrix is extracted as the detection feature. Thus, the detector based on the connected density of the graph is proposed for the floating small targets in sea surface. The analysis of the connected density of the measured Ice multiParameter Imaging X-band(IPIX) radar data shows that the graph composed by the sea clutter is relatively dense, whereas the graph composed by the floating small targets is relatively sparse. Thus, the connected density can effectively distinguish the floating small targets between the sea clutter. Furthermore, the experimental results show that, compared with other algorithms, the detection performance of the proposed connected density of the graph algorithm is obviously superior.
Due to the weak energy of the floating small targets, it is hard to be detected in sea surface. Relying on the energy, the traditional detectors based on statistical model inevitable loss the detection performance, regardless of the correlation between the frequency domain amplitudes. Therefore, in the paper, the correlation between the frequency domain amplitudes is considered by using the graph. Firstly, the connected density is calculated by the correlation between the frequency domain amplitudes of the echo pulses. Secondly, an adjacency matrix is generated based on the correlation. Thirdly, the adjacency matrix is converted to a Laplacian matrix. Lastly, the maximum eigenvalue of the Laplacian matrix is extracted as the detection feature. Thus, the detector based on the connected density of the graph is proposed for the floating small targets in sea surface. The analysis of the connected density of the measured Ice multiParameter Imaging X-band(IPIX) radar data shows that the graph composed by the sea clutter is relatively dense, whereas the graph composed by the floating small targets is relatively sparse. Thus, the connected density can effectively distinguish the floating small targets between the sea clutter. Furthermore, the experimental results show that, compared with other algorithms, the detection performance of the proposed connected density of the graph algorithm is obviously superior.
2021, 43(11): 3193-3200.
doi: 10.11999/JEIT201020
Abstract:
Reference signal reconstruction is one of the key technologies for signal processing of passive radar based on digital TV signals. The quality of the reconstructed signal affects directly the time-domain clutter suppression effect of the surveillance signal. To solve the problem that the reconstructed reference signal can not match the actual transmitted signal, this paper proposes a reference signal trusted reconstruction method based on “Demodulation-Remodulation” with the indicator that the optimal time-domain clutter suppression effect of the surveillance signal. First, the reference signal remodulation method is introduced and a signal model is established based on the non-ideal transmitted signal. Then, the theoretical relationship between the remodulation reference signal and the time-domain clutter suppression of the surveillance signal is derived. Based on the criterion of the optimal clutter suppression, the trusted reconstruction of the reference signal is obtained. Finally, simulation and field experiment verify the effectiveness of the reference signal trusted reconstruction method.
Reference signal reconstruction is one of the key technologies for signal processing of passive radar based on digital TV signals. The quality of the reconstructed signal affects directly the time-domain clutter suppression effect of the surveillance signal. To solve the problem that the reconstructed reference signal can not match the actual transmitted signal, this paper proposes a reference signal trusted reconstruction method based on “Demodulation-Remodulation” with the indicator that the optimal time-domain clutter suppression effect of the surveillance signal. First, the reference signal remodulation method is introduced and a signal model is established based on the non-ideal transmitted signal. Then, the theoretical relationship between the remodulation reference signal and the time-domain clutter suppression of the surveillance signal is derived. Based on the criterion of the optimal clutter suppression, the trusted reconstruction of the reference signal is obtained. Finally, simulation and field experiment verify the effectiveness of the reference signal trusted reconstruction method.
2021, 43(11): 3201-3210.
doi: 10.11999/JEIT200978
Abstract:
Considering the issue of poor target detection performance of millimeter-wave radar caused by the limited platform space and transmitting power in the case of autonomous driving, a joint design approach of waveform parameters and receiving weight is developed in this paper to improve the target detection probability of millimeter wave radar with range and velocity resolution constraints. Firstly, based on the Frequency Modulated Continuous Wave (FMCW) signal, the millimeter-wave phased array detection model is established via the proposed method; Secondly, the constraints of the transmitting waveform parameters concerning the range and velocity resolution are constructed by analyzing the relationship between the range along with speed resolution and the transmitting waveform parameters; After that, based on the criterion of maximizing the output Signal to Clutter plus Noise Ratio (SCNR), a joint optimization model of transmitting waveform parameters and receiving weight with range and velocity resolution constraints is established to improve the target detection and range-velocity resolution performance of millimeter wave radar; Finally, based on the alternate iteration method, the resultant complex nonlinear optimization problem can be solved via the developed approach. Simulation results show that the proposed method can adaptively adjust the transmitting waveform parameters and receiving weight to improve the target detection performance with satisfying the requirements of range and speed resolution.
Considering the issue of poor target detection performance of millimeter-wave radar caused by the limited platform space and transmitting power in the case of autonomous driving, a joint design approach of waveform parameters and receiving weight is developed in this paper to improve the target detection probability of millimeter wave radar with range and velocity resolution constraints. Firstly, based on the Frequency Modulated Continuous Wave (FMCW) signal, the millimeter-wave phased array detection model is established via the proposed method; Secondly, the constraints of the transmitting waveform parameters concerning the range and velocity resolution are constructed by analyzing the relationship between the range along with speed resolution and the transmitting waveform parameters; After that, based on the criterion of maximizing the output Signal to Clutter plus Noise Ratio (SCNR), a joint optimization model of transmitting waveform parameters and receiving weight with range and velocity resolution constraints is established to improve the target detection and range-velocity resolution performance of millimeter wave radar; Finally, based on the alternate iteration method, the resultant complex nonlinear optimization problem can be solved via the developed approach. Simulation results show that the proposed method can adaptively adjust the transmitting waveform parameters and receiving weight to improve the target detection performance with satisfying the requirements of range and speed resolution.
2021, 43(11): 3211-3219.
doi: 10.11999/JEIT210147
Abstract:
The available deception anti-jamming methods in multistatic radar can not be applied to multiple jamming sources. In view of this problem, a clustering analysis-based method is proposed to discriminate active false targets in multistatic radar, according to the differences in the correlation between the received signal vectors of true and false targets. Measured by the correlation coefficient in clustering analysis method, the false targets generated by the same jamming source are grouped into one cluster, while each true target constitutes one singleton cluster, achieving the effective discrimination of active false targets. The proposed method can be used to discriminate active false targets generated by any deception modulation mode and is suitable for the application of distributed jamming with multiple jamming sources. Finally, the effectiveness of the clustering analysis method to discriminate active false targets is verified through simulation experiments.
The available deception anti-jamming methods in multistatic radar can not be applied to multiple jamming sources. In view of this problem, a clustering analysis-based method is proposed to discriminate active false targets in multistatic radar, according to the differences in the correlation between the received signal vectors of true and false targets. Measured by the correlation coefficient in clustering analysis method, the false targets generated by the same jamming source are grouped into one cluster, while each true target constitutes one singleton cluster, achieving the effective discrimination of active false targets. The proposed method can be used to discriminate active false targets generated by any deception modulation mode and is suitable for the application of distributed jamming with multiple jamming sources. Finally, the effectiveness of the clustering analysis method to discriminate active false targets is verified through simulation experiments.
2021, 43(11): 3220-3227.
doi: 10.11999/JEIT201043
Abstract:
The Foreign Objects Debris (FOD) detection technology based on millimeter wave radar has the advantages of high resolution and low power consumption, but the traditional Constant False Alarm Rate (CFAR) detection algorithm has high false alarm probability under the condition of low Signal-to-Clutter Ratio (SCR). A FOD detection method based on Duffing oscillator is proposed. In this method, the clutter map CFAR detection method is firstly used to separate the background clutter from the received echo signal in the radar receiver, after that the distance information of target (including false alarm) can be acquired, and the Duffing equations are constructed by using the distance information. Then the Duffing equations are used as the system detection model, and the received echo signal is considered as the input. Therefore, the output variance can be calculated by solving the Duffing equations. Finally the target can be distinguished from the false alarm by using the variance extremum method. Simulation results show that, even if the false alarm probability is 10–3, the detection method in this paper can distinguish the target from the false alarm automatically under the condition of low SCR. Furthermore, it can also reduce the false alarm probability. Compared with the traditional CFAR detection algorithm, the detection probability of this method is higher and reduces more slowly with the decrease of SCR. Meanwhile, the detection probability can be maintained at 84% under the condition of SCR=–30 dB.
The Foreign Objects Debris (FOD) detection technology based on millimeter wave radar has the advantages of high resolution and low power consumption, but the traditional Constant False Alarm Rate (CFAR) detection algorithm has high false alarm probability under the condition of low Signal-to-Clutter Ratio (SCR). A FOD detection method based on Duffing oscillator is proposed. In this method, the clutter map CFAR detection method is firstly used to separate the background clutter from the received echo signal in the radar receiver, after that the distance information of target (including false alarm) can be acquired, and the Duffing equations are constructed by using the distance information. Then the Duffing equations are used as the system detection model, and the received echo signal is considered as the input. Therefore, the output variance can be calculated by solving the Duffing equations. Finally the target can be distinguished from the false alarm by using the variance extremum method. Simulation results show that, even if the false alarm probability is 10–3, the detection method in this paper can distinguish the target from the false alarm automatically under the condition of low SCR. Furthermore, it can also reduce the false alarm probability. Compared with the traditional CFAR detection algorithm, the detection probability of this method is higher and reduces more slowly with the decrease of SCR. Meanwhile, the detection probability can be maintained at 84% under the condition of SCR=–30 dB.
2021, 43(11): 3228-3236.
doi: 10.11999/JEIT200714
Abstract:
In the problem of sub-Nyquist sampled pulse Doppler radar signals, the existing methods have poor anti-noise performance, and the subsequent parameter estimation in the sequential parameter estimation methods is seriously affected by the accuracy of the previous parameter estimation. A Frequency-domain Delay-Doppler Two-dimensional Focusing (FD2TF) algorithm is proposed based on Finite Rate of Innovation (FRI) sampling method to solve the problem. The algorithm can obtain a series of Fourier coefficients of the signal at a sampling rate lower than the Nyquist sampling frequency through the FRI sampling structure. The time delay and Doppler parameters can be estimated simultaneously through the frequency-domain two-dimensional focusing process, and the problem of error accumulation in parameter sequential estimation methods can be avoided. Theoretical analysis proves that the algorithm can greatly improve the signal-to-noise ratio of the sampled signal, and improve the anti-noise performance and robustness of the algorithm. This paper also proposes a two-dimensional focusing simplification algorithm based on inverse Fourier transform, which greatly reduces the computational complexity of the two-dimensional focusing algorithm while increasing the grid density of parameter estimation. Simulation and comparative experiment results show that the proposed method is effective and has good anti-noise performance.
In the problem of sub-Nyquist sampled pulse Doppler radar signals, the existing methods have poor anti-noise performance, and the subsequent parameter estimation in the sequential parameter estimation methods is seriously affected by the accuracy of the previous parameter estimation. A Frequency-domain Delay-Doppler Two-dimensional Focusing (FD2TF) algorithm is proposed based on Finite Rate of Innovation (FRI) sampling method to solve the problem. The algorithm can obtain a series of Fourier coefficients of the signal at a sampling rate lower than the Nyquist sampling frequency through the FRI sampling structure. The time delay and Doppler parameters can be estimated simultaneously through the frequency-domain two-dimensional focusing process, and the problem of error accumulation in parameter sequential estimation methods can be avoided. Theoretical analysis proves that the algorithm can greatly improve the signal-to-noise ratio of the sampled signal, and improve the anti-noise performance and robustness of the algorithm. This paper also proposes a two-dimensional focusing simplification algorithm based on inverse Fourier transform, which greatly reduces the computational complexity of the two-dimensional focusing algorithm while increasing the grid density of parameter estimation. Simulation and comparative experiment results show that the proposed method is effective and has good anti-noise performance.
2021, 43(11): 3237-3246.
doi: 10.11999/JEIT201045
Abstract:
The existing target depth classification methods based on acoustic interference structure characteristics have a limited frequency range, and are only suitable for the case where target frequency can excite first two modes. Aiming to this problem, a novel target depth classification algorithm based on matching field processing is proposed in this paper. The proposed algorithm uses the reactive component of vertical complex acoustic intensity as matching variable to estimate target depth. The rough estimation results of target depth can be used to assist the binary classification of target depth. This method is suitable for the case where target frequency can excite first three modes, so as to expand the frequency range of the algorithm. The feasibility and robustness of the algorithm are verified by simulation results. The influence of mismatched sea environmental parameters on algorithm performance are also analyzed. The proposed algorithm has high accuracy and robustness.
The existing target depth classification methods based on acoustic interference structure characteristics have a limited frequency range, and are only suitable for the case where target frequency can excite first two modes. Aiming to this problem, a novel target depth classification algorithm based on matching field processing is proposed in this paper. The proposed algorithm uses the reactive component of vertical complex acoustic intensity as matching variable to estimate target depth. The rough estimation results of target depth can be used to assist the binary classification of target depth. This method is suitable for the case where target frequency can excite first three modes, so as to expand the frequency range of the algorithm. The feasibility and robustness of the algorithm are verified by simulation results. The influence of mismatched sea environmental parameters on algorithm performance are also analyzed. The proposed algorithm has high accuracy and robustness.
2021, 43(11): 3247-3256.
doi: 10.11999/JEIT201080
Abstract:
Compared with traditional swarm intelligence optimization algorithms, the Whale Optimization Algorithm(WOA) has better optimization capabilities and robustness, but there are still problems such as limited global optimization capabilities and difficulty in jumping out of local extremes. Considering the above-mentioned imbalance problem, a multi-group population division idea with vertical and horizontal bidirectional learning is proposed. The subgroups are independent of each other, and the individuals in the subgroups are affected by the optimal values from both the horizontal and vertical directions, thereby avoiding the local optimal and getting the balance between exploration and development.For all individuals in the vertical population, an individual replacement strategy with linearly decreasing probability is proposed to promote the information flow of different subgroups and accelerate the algorithm convergence.The selection of strategy operators is based on the historical evolution information of different individuals, which is different from the existing strategy operator selection methods based on random numbers.The benchmark function is used for cross-document comparison. The numerical results show that the algorithm in this thesis has good superiority and stability. It obtains global extreme on most problems and has good problem applicability.
Compared with traditional swarm intelligence optimization algorithms, the Whale Optimization Algorithm(WOA) has better optimization capabilities and robustness, but there are still problems such as limited global optimization capabilities and difficulty in jumping out of local extremes. Considering the above-mentioned imbalance problem, a multi-group population division idea with vertical and horizontal bidirectional learning is proposed. The subgroups are independent of each other, and the individuals in the subgroups are affected by the optimal values from both the horizontal and vertical directions, thereby avoiding the local optimal and getting the balance between exploration and development.For all individuals in the vertical population, an individual replacement strategy with linearly decreasing probability is proposed to promote the information flow of different subgroups and accelerate the algorithm convergence.The selection of strategy operators is based on the historical evolution information of different individuals, which is different from the existing strategy operator selection methods based on random numbers.The benchmark function is used for cross-document comparison. The numerical results show that the algorithm in this thesis has good superiority and stability. It obtains global extreme on most problems and has good problem applicability.
2021, 43(11): 3257-3265.
doi: 10.11999/JEIT201047
Abstract:
With the development of human-computer interaction, gesture recognition is becoming more and more important. At the same time, mobile terminal applications are developing rapidly, it is a development trend to implement human-computer interaction technology on the mobile terminal. An improved YOLOv4-tiny gesture recognition algorithm is proposed. Firstly, on the basis of YOLOv4-tiny network, the Spatial Pyramid Pooling(SPP) module is added to integrate the local and global features of the image to enhance the accurate positioning ability of the network. Secondly, a 1×1 convolution is added after the 3 maximum pooling layers of the original YOLOv4-tiny network and the newly added SPP module, which reduces the network parameters and improves the prediction speed of the network. On this basis, the K-means++ algorithm is used to generate an anchor box suitable for detecting gestures to speed up the network detection of gestures. In the gesture dataset NUS-II, compared with the YOLOv3-tiny algorithm and the YOLOv4-tiny algorithm, the improved algorithm mean Average Precision(mAP) is 100%, frames per second (fps) is 377, which can detect and recognize gestures quickly and accurately. The improved algorithm of this paper is deployed on the Android mobile terminal to realize the real-time gesture detection and recognition on the mobile terminal, which has great research significance for the development of human-computer interaction.
With the development of human-computer interaction, gesture recognition is becoming more and more important. At the same time, mobile terminal applications are developing rapidly, it is a development trend to implement human-computer interaction technology on the mobile terminal. An improved YOLOv4-tiny gesture recognition algorithm is proposed. Firstly, on the basis of YOLOv4-tiny network, the Spatial Pyramid Pooling(SPP) module is added to integrate the local and global features of the image to enhance the accurate positioning ability of the network. Secondly, a 1×1 convolution is added after the 3 maximum pooling layers of the original YOLOv4-tiny network and the newly added SPP module, which reduces the network parameters and improves the prediction speed of the network. On this basis, the K-means++ algorithm is used to generate an anchor box suitable for detecting gestures to speed up the network detection of gestures. In the gesture dataset NUS-II, compared with the YOLOv3-tiny algorithm and the YOLOv4-tiny algorithm, the improved algorithm mean Average Precision(mAP) is 100%, frames per second (fps) is 377, which can detect and recognize gestures quickly and accurately. The improved algorithm of this paper is deployed on the Android mobile terminal to realize the real-time gesture detection and recognition on the mobile terminal, which has great research significance for the development of human-computer interaction.
2021, 43(11): 3266-3271.
doi: 10.11999/JEIT200917
Abstract:
Recently in speaker recognition tasks, consistent performance gains have been continually achieved by various Convolutional Neural Networks (CNNs), which have shown increasingly stronger multiscale representation abilities. However, most existing methods enhance their strength with more layers and deeper structures. In this paper, a unique multiscale backbone architecture, Res2Net, is introduced for speaker recognition tasks, and its blocks are modified for assessment. This architecture works at a more granular level than most layer-wise networks. It improves the system by combining many equivalent receptive fields, resulting in a combination of different feature scales. The experiments results demonstrate that this architecture steadily achieves a 20% improvement on the Equal Error Rate (EER) over the baseline without additional computational burden. Its effectiveness and robustness are also verified in different environments and tasks, such as VoxCeleb and Speakers In The Wild (SITW). The modified full-connection block can make sure a more sufficient use of information and improves the performance obviously in more complex tasks. The code is available athttps://github.com/czg0326/Res2Net-Speaker-Recognition .
Recently in speaker recognition tasks, consistent performance gains have been continually achieved by various Convolutional Neural Networks (CNNs), which have shown increasingly stronger multiscale representation abilities. However, most existing methods enhance their strength with more layers and deeper structures. In this paper, a unique multiscale backbone architecture, Res2Net, is introduced for speaker recognition tasks, and its blocks are modified for assessment. This architecture works at a more granular level than most layer-wise networks. It improves the system by combining many equivalent receptive fields, resulting in a combination of different feature scales. The experiments results demonstrate that this architecture steadily achieves a 20% improvement on the Equal Error Rate (EER) over the baseline without additional computational burden. Its effectiveness and robustness are also verified in different environments and tasks, such as VoxCeleb and Speakers In The Wild (SITW). The modified full-connection block can make sure a more sufficient use of information and improves the performance obviously in more complex tasks. The code is available at
2021, 43(11): 3272-3281.
doi: 10.11999/JEIT210032
Abstract:
Hill climbing algorithm has too large search space and is easy to fall into local optimum. In this paper, a new Bayesian network structure algorithm based on V-structure & log-likelihood orientation and Tabu Hill (VTH) climbing is proposed. The algorithm limits the search space by using the oriented maximum weight spanning tree. In the process of maximum weight spanning tree orientation, the orientation strategy based on V-structure and Log-Likelihood (VLL) function is proposed. Tabu Hill Climbing (THC) scoring search strategy is established during the process of search, it combines the tabu list clearing mechanism with the local optimization criteria of hill climbing, the strategy not only ensures the search efficiency, but also improves the global optimization ability. By comparing Hamming distance, F1-value, Balanced Scoring Function(BSF) value and Time with other algorithms in Asia, Car, Child and Alarm standard networks, the effectiveness of the proposed algorithm is verified.
Hill climbing algorithm has too large search space and is easy to fall into local optimum. In this paper, a new Bayesian network structure algorithm based on V-structure & log-likelihood orientation and Tabu Hill (VTH) climbing is proposed. The algorithm limits the search space by using the oriented maximum weight spanning tree. In the process of maximum weight spanning tree orientation, the orientation strategy based on V-structure and Log-Likelihood (VLL) function is proposed. Tabu Hill Climbing (THC) scoring search strategy is established during the process of search, it combines the tabu list clearing mechanism with the local optimization criteria of hill climbing, the strategy not only ensures the search efficiency, but also improves the global optimization ability. By comparing Hamming distance, F1-value, Balanced Scoring Function(BSF) value and Time with other algorithms in Asia, Car, Child and Alarm standard networks, the effectiveness of the proposed algorithm is verified.
2021, 43(11): 3282-3290.
doi: 10.11999/JEIT210036
Abstract:
At present, aspect-level sentiment analysis uses mainly the method of combining attention mechanism and traditional neural network to model aspect and contextual words.These methods ignore the syntactic dependency information and position information between aspects and contextual words in sentences, which leads to unreasonable weight allocation of attention. Therefore, an Interactive Graph ATtention (IGATs) networks model for aspect-level sentiment analysis is proposed. Bidirectional Long Short-Term Memory (BiLSTM) network is firstly used to learn the semantic feature representation of sentences. And then the position information is combined to update the feature representation, a graph attention network is constructed on the newly generated feature representation to capture syntactic dependency information. Finally, interactive attention mechanism is used to model the semantic relations between the aspect and contextual words. Experimental results on three public datasets show that the accuracy and macro average F1 value of IGATs are significantly improved compared with other existing models.
At present, aspect-level sentiment analysis uses mainly the method of combining attention mechanism and traditional neural network to model aspect and contextual words.These methods ignore the syntactic dependency information and position information between aspects and contextual words in sentences, which leads to unreasonable weight allocation of attention. Therefore, an Interactive Graph ATtention (IGATs) networks model for aspect-level sentiment analysis is proposed. Bidirectional Long Short-Term Memory (BiLSTM) network is firstly used to learn the semantic feature representation of sentences. And then the position information is combined to update the feature representation, a graph attention network is constructed on the newly generated feature representation to capture syntactic dependency information. Finally, interactive attention mechanism is used to model the semantic relations between the aspect and contextual words. Experimental results on three public datasets show that the accuracy and macro average F1 value of IGATs are significantly improved compared with other existing models.
2021, 43(11): 3291-3300.
doi: 10.11999/JEIT201096
Abstract:
In order to meet the requirement of low resource cost and mixed noise suppression for outdoor target detection based on rotor Unmanned Aerial Vehicle (UAV), a mixed noise suppression algorithm based on Developable Local Surface (DLS) is proposed. This algorithm realizes the complementary advantages of the developable local surface algorithm and the layered noise reduction algorithm, and achieves the noise reduction effect that the neither algorithm can reach. Firstly, the developable local surface of image is used to suppress salt & pepper noise and low-density Gaussian noise in the image to obtain a preliminary denoised image. Then, the layered noise reduction in the spatial domain and the Fourier domain is carried, removing Gaussian noise and maximize the preservation of image edges, textures and other details. Finally, iteratively developable local surface and layered noise reduction to remove further residual components of mixed noise to achieve the purpose of suppressing mixed noise in target detection images. The experimental results show that the proposed algorithm has certain advantages over the other seven algorithms in removing mixed noise, and its subjective visual index and objective data index statistics are superior to those of the other seven algorithms.
In order to meet the requirement of low resource cost and mixed noise suppression for outdoor target detection based on rotor Unmanned Aerial Vehicle (UAV), a mixed noise suppression algorithm based on Developable Local Surface (DLS) is proposed. This algorithm realizes the complementary advantages of the developable local surface algorithm and the layered noise reduction algorithm, and achieves the noise reduction effect that the neither algorithm can reach. Firstly, the developable local surface of image is used to suppress salt & pepper noise and low-density Gaussian noise in the image to obtain a preliminary denoised image. Then, the layered noise reduction in the spatial domain and the Fourier domain is carried, removing Gaussian noise and maximize the preservation of image edges, textures and other details. Finally, iteratively developable local surface and layered noise reduction to remove further residual components of mixed noise to achieve the purpose of suppressing mixed noise in target detection images. The experimental results show that the proposed algorithm has certain advantages over the other seven algorithms in removing mixed noise, and its subjective visual index and objective data index statistics are superior to those of the other seven algorithms.
2021, 43(11): 3301-3308.
doi: 10.11999/JEIT201029
Abstract:
It needs a large number of training sets with annotation information to classify microstructure images of steel materials by deep learning. To solve the problem of low efficiency of manual image annotation, a new semi-supervised learning method combining self-organizing incremental neural network and graph convolutional neural network is proposed. Firstly, it uses transfer learning to obtain the feature vector set of images. Secondly, it obtains the topology structure by adopting the Weighted Self-Organizing Incremental Neural Network(WSOINN) based on connection weight strategy to learn feature data, and manually annotates a small number of nodes which are selected by the number of victories of node. Then, a Graph Convolution Network (GCN) is built to mine the potential connections of nodes in the graph, dropout is used to improve the generalization ability of the network, and the remaining nodes are automatically annotated to obtain the classification results of the metallograph. Experiment on the metallographic data collected from a state key laboratory, the accuracy of automatic classification under different manual annotation ratio is compared. The results show when the image annotation amount is only 12% of the traditional model, and the classification accuracy of the proposed model can reach up to 91%.
It needs a large number of training sets with annotation information to classify microstructure images of steel materials by deep learning. To solve the problem of low efficiency of manual image annotation, a new semi-supervised learning method combining self-organizing incremental neural network and graph convolutional neural network is proposed. Firstly, it uses transfer learning to obtain the feature vector set of images. Secondly, it obtains the topology structure by adopting the Weighted Self-Organizing Incremental Neural Network(WSOINN) based on connection weight strategy to learn feature data, and manually annotates a small number of nodes which are selected by the number of victories of node. Then, a Graph Convolution Network (GCN) is built to mine the potential connections of nodes in the graph, dropout is used to improve the generalization ability of the network, and the remaining nodes are automatically annotated to obtain the classification results of the metallograph. Experiment on the metallographic data collected from a state key laboratory, the accuracy of automatic classification under different manual annotation ratio is compared. The results show when the image annotation amount is only 12% of the traditional model, and the classification accuracy of the proposed model can reach up to 91%.
2021, 43(11): 3309-3318.
doi: 10.11999/JEIT201094
Abstract:
Capsule network is a new type of network model which is different from convolutional neural network. This paper attempts to improve its generalization and accuracy. Firstly, variational routing is used to alleviate the problem of classic routing that is highly dependent on prior information and can easily lead to model overfitting. By using the Gaussian Mixture Model (GMM) to fit the low-level matrix capsule and using the variational method to fit the approximation distribution, the error of the maximum likelihood point estimation is avoided, and the confidence calculation is used to improve the generalization performance; Secondly, considering that the actual data is mostly untagged or difficult to label, a capsule autoencoder with mutual information evaluation criterion is constructed to achieve effective selection of feature parameters. That is, by introducing a local encoder, only the most effective features in the capsule for identifying and classifying the original input are retained, which reduces the computational burden of the network while improving the accuracy of classification and recognition at the same time. The method in this paper is compared and tested on datasets such as MNIST, FashionMNIST, CIFAR-10, and CIFAR-100. The experimental results show that the performance of the proposed method is significantly improved compared with the classic capsule network.
Capsule network is a new type of network model which is different from convolutional neural network. This paper attempts to improve its generalization and accuracy. Firstly, variational routing is used to alleviate the problem of classic routing that is highly dependent on prior information and can easily lead to model overfitting. By using the Gaussian Mixture Model (GMM) to fit the low-level matrix capsule and using the variational method to fit the approximation distribution, the error of the maximum likelihood point estimation is avoided, and the confidence calculation is used to improve the generalization performance; Secondly, considering that the actual data is mostly untagged or difficult to label, a capsule autoencoder with mutual information evaluation criterion is constructed to achieve effective selection of feature parameters. That is, by introducing a local encoder, only the most effective features in the capsule for identifying and classifying the original input are retained, which reduces the computational burden of the network while improving the accuracy of classification and recognition at the same time. The method in this paper is compared and tested on datasets such as MNIST, FashionMNIST, CIFAR-10, and CIFAR-100. The experimental results show that the performance of the proposed method is significantly improved compared with the classic capsule network.
2021, 43(11): 3319-3326.
doi: 10.11999/JEIT200963
Abstract:
Early diagnosis of dementia is critical for timely treatment and intervention. Alzheimer’s Disease(AD) classification is an effective method on identifying AD at its early stage. In this paper, a feature selection method using improved Gauss fuzzy logic is proposed. Firstly, the normalized feature importance scores are calculated utilizing mutual information and variance analysis respectively. Then the final feature importance score is obtained by using improved Gauss fuzzy logic. At last, the features for AD classification are selected in accordance with the feature importance score. Furthermore, the heterogeneous ensemble classifier is constructed to classify AD patient utilizing selected features, which using logistic regression, random forest, LightGBM, support vector machine and depth feedforward network as primary classifier and multinomial naive Bayes classifier as secondary classifier. The proposed AD classification method is evaluated on the TADPOLE dataset. The experimental results show that the proposed feature selection method is effective and the integrated classifier based on Bayesian fusion is better than other conventional classification model on AD classification using the proposed feature selection method.
Early diagnosis of dementia is critical for timely treatment and intervention. Alzheimer’s Disease(AD) classification is an effective method on identifying AD at its early stage. In this paper, a feature selection method using improved Gauss fuzzy logic is proposed. Firstly, the normalized feature importance scores are calculated utilizing mutual information and variance analysis respectively. Then the final feature importance score is obtained by using improved Gauss fuzzy logic. At last, the features for AD classification are selected in accordance with the feature importance score. Furthermore, the heterogeneous ensemble classifier is constructed to classify AD patient utilizing selected features, which using logistic regression, random forest, LightGBM, support vector machine and depth feedforward network as primary classifier and multinomial naive Bayes classifier as secondary classifier. The proposed AD classification method is evaluated on the TADPOLE dataset. The experimental results show that the proposed feature selection method is effective and the integrated classifier based on Bayesian fusion is better than other conventional classification model on AD classification using the proposed feature selection method.
2021, 43(11): 3327-3334.
doi: 10.11999/JEIT201008
Abstract:
Using the methods of Optical Character Recognition (OCR) to detect and recognize the seal characters can speed up the classification speed and identification efficiency of all kinds of contracts. According to the characteristics of the cycle seal characters arranged in a ring, polar coordinate conversion is used to preprocess the seal characters, which overcomes the problem that the direction of the seal characters is not uniform. The Connectionist Text Proposal Network (CTPN) with angle information is used to detect the undulating text area, and the Bezier curve is used to achieve the accurate detection of the seal area. Finally, a method combined with the attention mechanism and the matching algorithm is used to recognize the detected text area and the seal text content is obtained. Using this algorithm to test the self-made Chinese seal data set, the F-measure of the seal content can reach 84.73%, and the recall rate of the character recognition is 84.4%, which shows that this algorithm can detect and recognize the seal content effectively, and has an important meaning for the research of document classification and identification.
Using the methods of Optical Character Recognition (OCR) to detect and recognize the seal characters can speed up the classification speed and identification efficiency of all kinds of contracts. According to the characteristics of the cycle seal characters arranged in a ring, polar coordinate conversion is used to preprocess the seal characters, which overcomes the problem that the direction of the seal characters is not uniform. The Connectionist Text Proposal Network (CTPN) with angle information is used to detect the undulating text area, and the Bezier curve is used to achieve the accurate detection of the seal area. Finally, a method combined with the attention mechanism and the matching algorithm is used to recognize the detected text area and the seal text content is obtained. Using this algorithm to test the self-made Chinese seal data set, the F-measure of the seal content can reach 84.73%, and the recall rate of the character recognition is 84.4%, which shows that this algorithm can detect and recognize the seal content effectively, and has an important meaning for the research of document classification and identification.
2021, 43(11): 3335-3341.
doi: 10.11999/JEIT200708
Abstract:
Object tracking based on structured SVM attracts much attention due to its excellent performance. However, the existing methods have the problem of imbalance between positive and negative samples. To solve the problem, a cost sensitive structured SVM model is proposed for object tracking. Secondly, an algorithm for the proposed model is designed via dual coordinate descent principle. Finally, a multi-scale object tracking method is implemented using the proposed cost sensitive structured SVM. The experimental results on OTB100 datasets and VOT2019 datasets show that compared with the correlation filtering trackers, the proposed method has higher tracking accuracy, and has the advantage of speed compared with the deep object trackers.
Object tracking based on structured SVM attracts much attention due to its excellent performance. However, the existing methods have the problem of imbalance between positive and negative samples. To solve the problem, a cost sensitive structured SVM model is proposed for object tracking. Secondly, an algorithm for the proposed model is designed via dual coordinate descent principle. Finally, a multi-scale object tracking method is implemented using the proposed cost sensitive structured SVM. The experimental results on OTB100 datasets and VOT2019 datasets show that compared with the correlation filtering trackers, the proposed method has higher tracking accuracy, and has the advantage of speed compared with the deep object trackers.
2021, 43(11): 3342-3349.
doi: 10.11999/JEIT210154
Abstract:
As a traditional feature extraction algorithm, Canonical Correlation Analysis (CCA) has been excellently used for the field of pattern recognition. It aims to find the projection direction that makes the maximum of the correlation between two groups of modal data. However, since the algorithm is an unsupervised linear method, it can not use intrinsic geometry structures and supervised information hidden in data, which will cause difficulty in dealing with high-dimensional nonlinear data. Therefore, this paper proposes a new nonlinear feature extraction algorithm, namely Graph Enhanced Canonical Correlation Analysis (GECCA). The algorithm uses different components of the data to construct multiple component graphs, which retains effectively the complex manifold structures between the data. The algorithm utilizes the probability evaluation method to use class label information, and the graph enhancement method is utilized to integrate the geometry manifolds and the supervised information into the typical correlation analysis framework. Targeted experiments are designed on the face and handwritten digital image datasets to evaluate the algorithm. Good experimental results show the advantages of GECCA in image recognition.
As a traditional feature extraction algorithm, Canonical Correlation Analysis (CCA) has been excellently used for the field of pattern recognition. It aims to find the projection direction that makes the maximum of the correlation between two groups of modal data. However, since the algorithm is an unsupervised linear method, it can not use intrinsic geometry structures and supervised information hidden in data, which will cause difficulty in dealing with high-dimensional nonlinear data. Therefore, this paper proposes a new nonlinear feature extraction algorithm, namely Graph Enhanced Canonical Correlation Analysis (GECCA). The algorithm uses different components of the data to construct multiple component graphs, which retains effectively the complex manifold structures between the data. The algorithm utilizes the probability evaluation method to use class label information, and the graph enhancement method is utilized to integrate the geometry manifolds and the supervised information into the typical correlation analysis framework. Targeted experiments are designed on the face and handwritten digital image datasets to evaluate the algorithm. Good experimental results show the advantages of GECCA in image recognition.
2021, 43(11): 3350-3358.
doi: 10.11999/JEIT200946
Abstract:
Conditional Proxy Re-Encryption (CPRE) can grant fine-grained authorization to the original ciphertext according to the conditions. The existing CPRE schemes only check the conditions of the original ciphertext, but ignore the conditions of the re-encryption key. No effective measures are taken to protect the conditions in these CPRE schemes, which may lead to privacy disclosure of conditions. A Threshold-Based Conditional Anonymous Proxy Re-Encryption scheme (TB-CAPRE) is constructed, which can not only verify the conditions of ciphertext and re-encryption key at the same time, but also make sensitive conditional information anonymous. The re-encryption processes are completed by multiple agent nodes, so TB-CAPRE can resist the collusion attacks. The theoretical analysis proves that the new scheme is INDistinguishable against adaptive Chosen-Ciphertext Attack(simply donoted by IND-CCA) in the random oracle. The analysis of performance and computation shows that TB-CAPRE does not introduce excessive overhead while increasing security and related functions. It is possible that TB-CAPRE is applied to distributed environment.
Conditional Proxy Re-Encryption (CPRE) can grant fine-grained authorization to the original ciphertext according to the conditions. The existing CPRE schemes only check the conditions of the original ciphertext, but ignore the conditions of the re-encryption key. No effective measures are taken to protect the conditions in these CPRE schemes, which may lead to privacy disclosure of conditions. A Threshold-Based Conditional Anonymous Proxy Re-Encryption scheme (TB-CAPRE) is constructed, which can not only verify the conditions of ciphertext and re-encryption key at the same time, but also make sensitive conditional information anonymous. The re-encryption processes are completed by multiple agent nodes, so TB-CAPRE can resist the collusion attacks. The theoretical analysis proves that the new scheme is INDistinguishable against adaptive Chosen-Ciphertext Attack(simply donoted by IND-CCA) in the random oracle. The analysis of performance and computation shows that TB-CAPRE does not introduce excessive overhead while increasing security and related functions. It is possible that TB-CAPRE is applied to distributed environment.
2021, 43(11): 3359-3366.
doi: 10.11999/JEIT200999
Abstract:
SIMON algorithm is a group of lightweight block cipher algorithms introduced by the National Security Agency (NSA) in 2013. It has the advantages of low implementation cost and good security performance. Its round function adopts\begin{document}$F(x) = (x < < < a){{\& }}(x < < < b) \oplus (x < < < c)$\end{document} ![]()
![]()
\begin{document}${2^{ - k}}$\end{document} ![]()
![]()
\begin{document}$k \in Z$\end{document} ![]()
![]()
\begin{document}${{0}} \le k \le \left\lfloor {{2^{ - 1}}n} \right\rfloor $\end{document} ![]()
![]()
\begin{document}${2^{ - k}}$\end{document} ![]()
![]()
SIMON algorithm is a group of lightweight block cipher algorithms introduced by the National Security Agency (NSA) in 2013. It has the advantages of low implementation cost and good security performance. Its round function adopts
2021, 43(11): 3367-3373.
doi: 10.11999/JEIT200723
Abstract:
Although Deep Neural Networks (DNN) achieves high accuracy in image recognition, it is significantly vulnerable to adversarial examples. Adversarial training is one of the effective methods to resist adversarial examples empirically. Generating more powerful adversarial examples can solve the inner maximization problem of adversarial training better, which is the key to improve the effectiveness of adversarial training. In this paper, to solve the inner maximization problem, an adversarial training based on second-order adversarial examples is proposed to generate more powerful adversarial examples through quadratic polynomial approximation in a tiny input neighborhood. Through theoretical analysis, second-order adversarial examples are shown to outperform first-order adversarial examples. Experiments on MNIST and CIFAR10 data sets show that second-order adversarial examples have high attack success rate and high concealment. Compared with PGD adversarial training, adversarial training based on second-order adversarial examples is robust to all the existing typical attacks.
Although Deep Neural Networks (DNN) achieves high accuracy in image recognition, it is significantly vulnerable to adversarial examples. Adversarial training is one of the effective methods to resist adversarial examples empirically. Generating more powerful adversarial examples can solve the inner maximization problem of adversarial training better, which is the key to improve the effectiveness of adversarial training. In this paper, to solve the inner maximization problem, an adversarial training based on second-order adversarial examples is proposed to generate more powerful adversarial examples through quadratic polynomial approximation in a tiny input neighborhood. Through theoretical analysis, second-order adversarial examples are shown to outperform first-order adversarial examples. Experiments on MNIST and CIFAR10 data sets show that second-order adversarial examples have high attack success rate and high concealment. Compared with PGD adversarial training, adversarial training based on second-order adversarial examples is robust to all the existing typical attacks.
2021, 43(11): 3374-3383.
doi: 10.11999/JEIT210026
Abstract:
In this paper, a new locally active memristor is proposed. The characteristics and local activity of the memristor are analyzed by using standard nonlinear theory and the circuit theoretic technique direct current loci. Furthermore, the locally active memristor is used to simulate a biological synapse, then a locally active memristive synaptic coupled Hindmarsh-Rose (HR) neuron network is constructed. Theoretical analysis and numerical simulation show that the memristive neural network can generate multiple firing patterns and complex chaotic behaviors under the influence of locally active memristive synapses. Finally, the equivalent analog circuit of the memristive synaptic coupled neuron network is designed, and the correctness of numerical simulation is verified by Power SIMulation (PSIM) circuit simulations.
In this paper, a new locally active memristor is proposed. The characteristics and local activity of the memristor are analyzed by using standard nonlinear theory and the circuit theoretic technique direct current loci. Furthermore, the locally active memristor is used to simulate a biological synapse, then a locally active memristive synaptic coupled Hindmarsh-Rose (HR) neuron network is constructed. Theoretical analysis and numerical simulation show that the memristive neural network can generate multiple firing patterns and complex chaotic behaviors under the influence of locally active memristive synapses. Finally, the equivalent analog circuit of the memristive synaptic coupled neuron network is designed, and the correctness of numerical simulation is verified by Power SIMulation (PSIM) circuit simulations.
2021, 43(11): 3384-3392.
doi: 10.11999/JEIT201079
Abstract:
In this paper, a passive flux-controlled memristor is used to replace the diode series-parallel branch in the Shinriki oscillator, and the active flux-controlled memristor is introduced to substitute the resistance in the RLC resonant loop. At the same time, a series resistance is connected in the inductance branch to obtain a new type of dual flux-controlled memristive Shinriki oscillator. Through the coexisting bifurcation diagram of specific parameters and the Lyapunov exponential spectrum, the symmetric bifurcation behavior of oscillators is innovatively discovered, and the symmetry of the motion state distribution is shown in the two-parameter plane. Meanwhile, in the basin of attraction of the symmetrical parameter-initial value plane, the multistable characteristics of the system in the symmetrical domain are analyzed. The existence of symmetrical antimonotonic phenomena, the symmetrical coexistence of attractors with multiple motion states, and the incomplete symmetry behavior that depends on the initial value in the symmetric domain are studied. In addition, the digital circuit experiment of the dual flux-controlled memristive Shinriki oscillator is completed based on FPGA technology, and the waveform captured on the oscilloscope verifies the correctness of the system’s symmetrical dynamic behavior analysis.
In this paper, a passive flux-controlled memristor is used to replace the diode series-parallel branch in the Shinriki oscillator, and the active flux-controlled memristor is introduced to substitute the resistance in the RLC resonant loop. At the same time, a series resistance is connected in the inductance branch to obtain a new type of dual flux-controlled memristive Shinriki oscillator. Through the coexisting bifurcation diagram of specific parameters and the Lyapunov exponential spectrum, the symmetric bifurcation behavior of oscillators is innovatively discovered, and the symmetry of the motion state distribution is shown in the two-parameter plane. Meanwhile, in the basin of attraction of the symmetrical parameter-initial value plane, the multistable characteristics of the system in the symmetrical domain are analyzed. The existence of symmetrical antimonotonic phenomena, the symmetrical coexistence of attractors with multiple motion states, and the incomplete symmetry behavior that depends on the initial value in the symmetric domain are studied. In addition, the digital circuit experiment of the dual flux-controlled memristive Shinriki oscillator is completed based on FPGA technology, and the waveform captured on the oscilloscope verifies the correctness of the system’s symmetrical dynamic behavior analysis.
