Email alert
2018 Vol. 40, No. 11
Display Method:
2018, 40(11)
Abstract:
2018, 40(11): 2541-2546.
doi: 10.11999/JEIT180175
Abstract:
The structured random sampling strategy adopted in array diagnosis has negative influence on the performance of measurement matrix. Therefore, a compressed sensing based deterministic sampling strategy to diagnose defective array elements using far-field measurements is investigated in this paper. In the case of the number of failed elements satisfies sparsity, the sparse vector is constructed by subtracting incentives of reference array without failures and the array under test. Deterministic Partial Fourier Matrix (DPFM) is then formulated by the proposed strategy as the measurement matrix. Finally, accurate diagnosis with high probability is achieved by l1 norm minimization. Theoretical analysis and simulation results demonstrate that the proposed method can avoid the adverse impact on the performance of measurement matrix effectively arising from the random distribution of sampling positions, simplify the sampling procedure and improve the probability of success rate of diagnosis.
The structured random sampling strategy adopted in array diagnosis has negative influence on the performance of measurement matrix. Therefore, a compressed sensing based deterministic sampling strategy to diagnose defective array elements using far-field measurements is investigated in this paper. In the case of the number of failed elements satisfies sparsity, the sparse vector is constructed by subtracting incentives of reference array without failures and the array under test. Deterministic Partial Fourier Matrix (DPFM) is then formulated by the proposed strategy as the measurement matrix. Finally, accurate diagnosis with high probability is achieved by l1 norm minimization. Theoretical analysis and simulation results demonstrate that the proposed method can avoid the adverse impact on the performance of measurement matrix effectively arising from the random distribution of sampling positions, simplify the sampling procedure and improve the probability of success rate of diagnosis.
2018, 40(11): 2547-2553.
doi: 10.11999/JEIT180141
Abstract:
The traditional Total Variation (TV) model based on local operators for texture image colorization has some problems, such as inhomogeneous color diffusion, small coloring ranges and so on. In order to solve these problems, a coupled total variation model based on nonlocal operators is presented for image colorization, and the correspond numerical algorithm is designed to solve the model by incorporating the Alternating Direction Method of Multipliers (ADMM), and the convergence result of the algorithm is given. The proposed model makes full use of the similarity between the brightness of the pixel areas to perform color diffusion, which can effectively avoid the problem of inhomogeneous color diffusion due to local diffusion only using the brightness edge information. The experimental results are given to show that the model can effectively solve the problem of inhomogeneous color diffusion at textures and other details while fast colorizing.
The traditional Total Variation (TV) model based on local operators for texture image colorization has some problems, such as inhomogeneous color diffusion, small coloring ranges and so on. In order to solve these problems, a coupled total variation model based on nonlocal operators is presented for image colorization, and the correspond numerical algorithm is designed to solve the model by incorporating the Alternating Direction Method of Multipliers (ADMM), and the convergence result of the algorithm is given. The proposed model makes full use of the similarity between the brightness of the pixel areas to perform color diffusion, which can effectively avoid the problem of inhomogeneous color diffusion due to local diffusion only using the brightness edge information. The experimental results are given to show that the model can effectively solve the problem of inhomogeneous color diffusion at textures and other details while fast colorizing.
2018, 40(11): 2554-2561.
doi: 10.11999/JEIT180118
Abstract:
Traditional saliency object detection methods, assuming that there is only one salient object, is not conductive to practical application. Their effects are dependent on saliency threshold. Object detection model provides a kind of new solutions. SSD can accurately detect multi-objects with different scales simultaneously, except for small objects. To overcome this drawback, this paper presents a new multi- saliency objects detection model, DAR-SSD, appending a deconvolution module embedded with an attention residual module. Experiments show that DAR-SSD achieves a higher detection accuracy than SOD. Also, it improves detection performance for multi- saliency objects on small scales, compared with original SSD, and it has an advantage over complicated background, compared with MDF and DCL, which also are deep model based methods.
Traditional saliency object detection methods, assuming that there is only one salient object, is not conductive to practical application. Their effects are dependent on saliency threshold. Object detection model provides a kind of new solutions. SSD can accurately detect multi-objects with different scales simultaneously, except for small objects. To overcome this drawback, this paper presents a new multi- saliency objects detection model, DAR-SSD, appending a deconvolution module embedded with an attention residual module. Experiments show that DAR-SSD achieves a higher detection accuracy than SOD. Also, it improves detection performance for multi- saliency objects on small scales, compared with original SSD, and it has an advantage over complicated background, compared with MDF and DCL, which also are deep model based methods.
2018, 40(11): 2562-2569.
doi: 10.11999/JEIT171141
Abstract:
To solve the common problem of classification performance restriction caused by big intra-class variations and inter-class similarities in video classification domain, this paper proposes a deep metric learning based video classification method. The proposed method designs a deep network which contains three parts: feature learning, deep metric learning based similarity measure as well as classification. The principle of similarity measure is: Firstly, the Euclidean distance between features is calculated as the semantic distance between samples. Secondly, a margin distributing function is designed to dynamically allocate margin in the basis of the semantic distances. Finally, the difference of the sample semantic distance can be learned by calculating the loss and propagating it backwards so as to the network can automatically focus on the hard negative samples and more fully learn the characteristic of them. With a multi-task learning training method in the training stage, the similarity measure and classification can be learned jointly. Experimental results on UCF101 and HMDB51 show that the proposed method can effectively improve the classification precision.
To solve the common problem of classification performance restriction caused by big intra-class variations and inter-class similarities in video classification domain, this paper proposes a deep metric learning based video classification method. The proposed method designs a deep network which contains three parts: feature learning, deep metric learning based similarity measure as well as classification. The principle of similarity measure is: Firstly, the Euclidean distance between features is calculated as the semantic distance between samples. Secondly, a margin distributing function is designed to dynamically allocate margin in the basis of the semantic distances. Finally, the difference of the sample semantic distance can be learned by calculating the loss and propagating it backwards so as to the network can automatically focus on the hard negative samples and more fully learn the characteristic of them. With a multi-task learning training method in the training stage, the similarity measure and classification can be learned jointly. Experimental results on UCF101 and HMDB51 show that the proposed method can effectively improve the classification precision.
2018, 40(11): 2570-2577.
doi: 10.11999/JEIT171040
Abstract:
To solve the problem of the line spectrum estimation under colored noise background, a subband line spectrum estimation method using sparse reconstruction is proposed. Firstly, the input signal is divided into several subbands by a multi-rate cosine modulated filter bank. The subband signal has the flatter power spectrum. The sparse learning via iterative minimization method is utilized on each subband to estimate the line spectrum signal. Then, the results of line spectrum estimation on each subband are processed by frequency domain synthesis filtering and threshold decision. Finally, the line spectrum signal under colored noise background is identified. Theoretical derivation and simulation experiments show that the proposed method has better line spectrum estimation performance under colored noise background. The colored noise background can be removed, and the advantage of high frequency resolution of sparse reconstruction method is retained.
To solve the problem of the line spectrum estimation under colored noise background, a subband line spectrum estimation method using sparse reconstruction is proposed. Firstly, the input signal is divided into several subbands by a multi-rate cosine modulated filter bank. The subband signal has the flatter power spectrum. The sparse learning via iterative minimization method is utilized on each subband to estimate the line spectrum signal. Then, the results of line spectrum estimation on each subband are processed by frequency domain synthesis filtering and threshold decision. Finally, the line spectrum signal under colored noise background is identified. Theoretical derivation and simulation experiments show that the proposed method has better line spectrum estimation performance under colored noise background. The colored noise background can be removed, and the advantage of high frequency resolution of sparse reconstruction method is retained.
2018, 40(11): 2578-2583.
doi: 10.11999/JEIT180056
Abstract:
Complicated underwater environment puts forward high requirements on the fault-tolerant and reliability of underwater acoustic localization systems. An anti-outlier localization method based on K-Means Clustering and Decision Fusion (KMCDF) is proposed for integrated Long baseline/Ultra-Short BaseLine (L/USBL) systems. Firstly, the target position is preliminarily estimated by the multi-parameter redundant information measured by the integrated system. Then, the clustering degree of the preliminary coordinates is analyzed by k-means clustering. According to the incompatibility between outliers and normal parameters, the outliers are identified by the decision fusion method. Furthermore, the impact of outliers on positioning is eliminated. Simulation analysis shows that the proposed method fully incorporates the multi-parameter information, and the tolerance of outliers is better than the existing anti-outlier positioning methods based on the time-delay parameter. Lake trial results demonstrate further the effectiveness of the proposed method.
Complicated underwater environment puts forward high requirements on the fault-tolerant and reliability of underwater acoustic localization systems. An anti-outlier localization method based on K-Means Clustering and Decision Fusion (KMCDF) is proposed for integrated Long baseline/Ultra-Short BaseLine (L/USBL) systems. Firstly, the target position is preliminarily estimated by the multi-parameter redundant information measured by the integrated system. Then, the clustering degree of the preliminary coordinates is analyzed by k-means clustering. According to the incompatibility between outliers and normal parameters, the outliers are identified by the decision fusion method. Furthermore, the impact of outliers on positioning is eliminated. Simulation analysis shows that the proposed method fully incorporates the multi-parameter information, and the tolerance of outliers is better than the existing anti-outlier positioning methods based on the time-delay parameter. Lake trial results demonstrate further the effectiveness of the proposed method.
2018, 40(11): 2584-2589.
doi: 10.11999/JEIT180001
Abstract:
The Compressed Sensing (CS) Multiple Measurement Vector (MMV) model is used to solve multiple snapshots problem with the same sparse structure. MUltiple SIgnal Classification (MUSIC) is a common method in traditional array signal processing applications. However, when the number of snapshots is below sparsity performance will be dramatically deteriorated. Kim et al. derive a modified MUSIC spectral method and propose a Compressed Sensing MUSIC method (CS-MUSIC) combining the compression reconstruction method and the MUSIC algorithm, which can effectively overcome the problem of insufficient snapshot number. In this paper, Kim et al.’s conclusion is extended to the general case, and a Modified MUSIC (MMUSIC) algorithm is proposed based on the traditional MUSIC method and the CS-MUSIC method. The simulation results show that the proposed algorithm can effectively overcome the shortage of snapshots and has a higher reconstruction probability than the CS-MUSIC algorithm and the compressed sensing greedy algorithm.
The Compressed Sensing (CS) Multiple Measurement Vector (MMV) model is used to solve multiple snapshots problem with the same sparse structure. MUltiple SIgnal Classification (MUSIC) is a common method in traditional array signal processing applications. However, when the number of snapshots is below sparsity performance will be dramatically deteriorated. Kim et al. derive a modified MUSIC spectral method and propose a Compressed Sensing MUSIC method (CS-MUSIC) combining the compression reconstruction method and the MUSIC algorithm, which can effectively overcome the problem of insufficient snapshot number. In this paper, Kim et al.’s conclusion is extended to the general case, and a Modified MUSIC (MMUSIC) algorithm is proposed based on the traditional MUSIC method and the CS-MUSIC method. The simulation results show that the proposed algorithm can effectively overcome the shortage of snapshots and has a higher reconstruction probability than the CS-MUSIC algorithm and the compressed sensing greedy algorithm.
2018, 40(11): 2590-2597.
doi: 10.11999/JEIT171088
Abstract:
Recent researches show great interests in localizing dynamic objects through cost-effective technologies. Laser or visual-based approaches have to solve the singularity and occlusion problem from the environment. Radio Frequency IDentification (RFID) is used as a preferred technology to address these issues, due to the unique identification and the communication without line of sight. In this paper, an innovative method is proposed to localize precisely a dynamic object equipped with an RFID tag by fusing laser information RFID information. A particle filter is used to fuse RFID signal strength, phase information, and laser ranging data. Particularly, a pre-trained signal strength-based model is used to incorporate the signal strength information. Then, the laser ranging data is divided into different clusters and the velocities of these clusters are compared with the RFID phase velocity. Matching results of both velocities are used to confine the locations of the particles during the update stage of the particle filtering. The proposed approach is verified by several experiments on a SCITOS service robot and results show that the proposed approach provides better localization accuracy when compared with laser-based approach and the signal strength-based approach.
Recent researches show great interests in localizing dynamic objects through cost-effective technologies. Laser or visual-based approaches have to solve the singularity and occlusion problem from the environment. Radio Frequency IDentification (RFID) is used as a preferred technology to address these issues, due to the unique identification and the communication without line of sight. In this paper, an innovative method is proposed to localize precisely a dynamic object equipped with an RFID tag by fusing laser information RFID information. A particle filter is used to fuse RFID signal strength, phase information, and laser ranging data. Particularly, a pre-trained signal strength-based model is used to incorporate the signal strength information. Then, the laser ranging data is divided into different clusters and the velocities of these clusters are compared with the RFID phase velocity. Matching results of both velocities are used to confine the locations of the particles during the update stage of the particle filtering. The proposed approach is verified by several experiments on a SCITOS service robot and results show that the proposed approach provides better localization accuracy when compared with laser-based approach and the signal strength-based approach.
2018, 40(11): 2598-2605.
doi: 10.11999/JEIT180163
Abstract:
Face detection is finding and locating all faces in the input image, and then returning the position and size of the faces. It is an important direction of target detection. In order to solve the problem which is caused by the diversity of face size, a new single shot multiscale face algorithm is presented based on feature fusion. This method combines predictions from multiple feature maps with different resolutions to handle faces of various sizes, and the fusion of the feature maps in the shallow layers can improve the detection accuracy of the small size face by introducing the contextual information. Experimental results on the FDDB and WIDERFACE datasets confirm that the proposed method has competitive accuracy. Additionally, the object proposal step is removed, which makes the method fast. The proposed model achieves 87.9%, 93.2% and 93.4% Mean Average Precision (MAP) on the WIDERFACE sub-datasets respectively, at 35 fps. The proposed method outperforms a comparable state-of-the-art HR model, and at the same time improves the speed while ensuring the accuracy.
Face detection is finding and locating all faces in the input image, and then returning the position and size of the faces. It is an important direction of target detection. In order to solve the problem which is caused by the diversity of face size, a new single shot multiscale face algorithm is presented based on feature fusion. This method combines predictions from multiple feature maps with different resolutions to handle faces of various sizes, and the fusion of the feature maps in the shallow layers can improve the detection accuracy of the small size face by introducing the contextual information. Experimental results on the FDDB and WIDERFACE datasets confirm that the proposed method has competitive accuracy. Additionally, the object proposal step is removed, which makes the method fast. The proposed model achieves 87.9%, 93.2% and 93.4% Mean Average Precision (MAP) on the WIDERFACE sub-datasets respectively, at 35 fps. The proposed method outperforms a comparable state-of-the-art HR model, and at the same time improves the speed while ensuring the accuracy.
2018, 40(11): 2606-2613.
doi: 10.11999/JEIT180024
Abstract:
After analyzing the features of three measured data from the low-resolution radar system, corresponding to the helicopter, the propeller, and the turbojet, an algorithm is proposed by using multiple features to classify and recognize the aircraft targets. First, multiple features are extracted, including Doppler frequency shift, relative magnitude, waveform entropy of time and frequency domain, and time-frequency domain features from the measured data. Then, these features are utilized for classification purpose by means of the Support Vector Machine (SVM). Finally, owing to the symmetry and the width of time-frequency distributions of the returned signals between the helicopters with odd and even blades, a method is proposed to recognize of helicopter. The experimental results of measured data verify the effectivity of the proposed algorithms.
After analyzing the features of three measured data from the low-resolution radar system, corresponding to the helicopter, the propeller, and the turbojet, an algorithm is proposed by using multiple features to classify and recognize the aircraft targets. First, multiple features are extracted, including Doppler frequency shift, relative magnitude, waveform entropy of time and frequency domain, and time-frequency domain features from the measured data. Then, these features are utilized for classification purpose by means of the Support Vector Machine (SVM). Finally, owing to the symmetry and the width of time-frequency distributions of the returned signals between the helicopters with odd and even blades, a method is proposed to recognize of helicopter. The experimental results of measured data verify the effectivity of the proposed algorithms.
2018, 40(11): 2614-2620.
doi: 10.11999/JEIT180054
Abstract:
Under the condition of lack of echo data and low SNR, the ISAR imaging performance is greatly reduced by using Random Chirp Frequency-Stepped (RCFS) signal. To solve the above problems, based on fully analyzing the echo characteristics of the random chirp frequency-stepped signal, a new method of obtaining high quality ISAR images is proposed using the joint sparse feature of the target range dimension. First, a joint block sparse imaging model of the target echo signal under the condition of random chirp frequency-stepped signal is derived and the characteristics of the model are analyzed. Secondly, a Joint Block sparse Orthogonal Matching Pursuit (JBOMP) algorithm is proposed for solving the model. The algorithm utilizes the sparse information and the joint sparse information of the ISAR echo. Therefore, the ISAR imaging performance is enhanced under the condition of low measurement and low SNR. The proposed algorithm also can achieve joint processing of multidimensional signals and has a faster operation speed. Both theoretical analysis and simulation experiments verify the effectiveness of the proposed method.
Under the condition of lack of echo data and low SNR, the ISAR imaging performance is greatly reduced by using Random Chirp Frequency-Stepped (RCFS) signal. To solve the above problems, based on fully analyzing the echo characteristics of the random chirp frequency-stepped signal, a new method of obtaining high quality ISAR images is proposed using the joint sparse feature of the target range dimension. First, a joint block sparse imaging model of the target echo signal under the condition of random chirp frequency-stepped signal is derived and the characteristics of the model are analyzed. Secondly, a Joint Block sparse Orthogonal Matching Pursuit (JBOMP) algorithm is proposed for solving the model. The algorithm utilizes the sparse information and the joint sparse information of the ISAR echo. Therefore, the ISAR imaging performance is enhanced under the condition of low measurement and low SNR. The proposed algorithm also can achieve joint processing of multidimensional signals and has a faster operation speed. Both theoretical analysis and simulation experiments verify the effectiveness of the proposed method.
2018, 40(11): 2621-2629.
doi: 10.11999/JEIT180021
Abstract:
The diving SAR usually adopts the highly squinted mode and sub-aperture to satisfy the maneuvering and real-time processing. However, the existence of severe range-azimuth coupling, range-dependent squint angle and three-dimension velocity and acceleration leads to the space variance of range envelope and azimuth phase, which makes imagery unfocused seriously. To solve these problems, a Two-stage Frequency Filtering Algorithm (TsFFA) is proposed. After preprocessing, the First-stage Frequency Filtering (FsFF) factor is first introduced to correct azimuth-dependent Range Cell Migration (RCM) and realize the unified RCM correction. Furthermore, the Second-stage Frequency Filtering (SsFF) factor is adopted to equalize azimuth-dependent Doppler parameters and realize unified azimuth phase focused. Simulation results are presented to validate the effectiveness of the proposed approach.
The diving SAR usually adopts the highly squinted mode and sub-aperture to satisfy the maneuvering and real-time processing. However, the existence of severe range-azimuth coupling, range-dependent squint angle and three-dimension velocity and acceleration leads to the space variance of range envelope and azimuth phase, which makes imagery unfocused seriously. To solve these problems, a Two-stage Frequency Filtering Algorithm (TsFFA) is proposed. After preprocessing, the First-stage Frequency Filtering (FsFF) factor is first introduced to correct azimuth-dependent Range Cell Migration (RCM) and realize the unified RCM correction. Furthermore, the Second-stage Frequency Filtering (SsFF) factor is adopted to equalize azimuth-dependent Doppler parameters and realize unified azimuth phase focused. Simulation results are presented to validate the effectiveness of the proposed approach.
2018, 40(11): 2630-2637.
doi: 10.11999/JEIT180099
Abstract:
When using Inverse Synthetic Aperture Radar (ISAR) to observe the spinning targets, the range-Doppler time-varying characteristics of spinning target echo would lead to the inefficiency of traditional imaging methods. To solve this problem, a fast high-resolution imaging method based on distributed matching sparse representation model is proposed for wideband spinning targets imaging. Firstly, a distributed matching sparse representation model is constructed based on the sparsity of spinning target echo. Secondly, a Fast Distributed Simultaneous Multiple Orthogonal Matching Pursuit (FDSMOMP) algorithm is proposed for achieving the fast robust imaging of the spinning parts. The proposed algorithm can significantly improve the reconstruction efficiency by reducing the iteration times and computational complexity of each iteration. Additionally, in order to enhance the robustness of FDSMOMP, a related threshold is designed to suppress the false reconstruction. Finally, the mechanism of the presented method is analyzed theoretically, and it is proved that the high quality imaging result can still be obtained under the conditions of sub-Nyquist sampling and lower (SNR Signal Noise Ratio). Simulation results show the validation of the proposed method.
When using Inverse Synthetic Aperture Radar (ISAR) to observe the spinning targets, the range-Doppler time-varying characteristics of spinning target echo would lead to the inefficiency of traditional imaging methods. To solve this problem, a fast high-resolution imaging method based on distributed matching sparse representation model is proposed for wideband spinning targets imaging. Firstly, a distributed matching sparse representation model is constructed based on the sparsity of spinning target echo. Secondly, a Fast Distributed Simultaneous Multiple Orthogonal Matching Pursuit (FDSMOMP) algorithm is proposed for achieving the fast robust imaging of the spinning parts. The proposed algorithm can significantly improve the reconstruction efficiency by reducing the iteration times and computational complexity of each iteration. Additionally, in order to enhance the robustness of FDSMOMP, a related threshold is designed to suppress the false reconstruction. Finally, the mechanism of the presented method is analyzed theoretically, and it is proved that the high quality imaging result can still be obtained under the conditions of sub-Nyquist sampling and lower (SNR Signal Noise Ratio). Simulation results show the validation of the proposed method.
2018, 40(11): 2638-2644.
doi: 10.11999/JEIT170994
Abstract:
The traditional bistatic equivalent range model has low accuracy and make the traditional Polar Format Algorithm (PFA) are inapplicable in missile-borne bistatic Synthetic Aperture Radar (SAR) imaging with curved track due to the existing of three-axis velocity and acceleration. In addition, due to the existing of space-variant motion error introduced by acceleration, the traditional 2-D sub-block compensation method will cause the discontinuities between the image sub-blocks, thus affecting the subsequent image matching application. In view of these problems, this paper proposes a Back-Filtering PFA algorithm (BFPFA) which is based on the Improved Generalized Bistatic Equivalent Range Model (IGBERM). Constructing the combined compensating filter of space-variant phase error and geometric distortion, as well as reverse mapping interpolation, can realize the combined compensation of motion error, wavefront bending and geometric distortion in the process of oblique conversion, and obtain the SAR distance map without distortion, which is more conducive to the subsequent image matching applications. Finally, the simulations validate the effectiveness of the proposed algorithm.
The traditional bistatic equivalent range model has low accuracy and make the traditional Polar Format Algorithm (PFA) are inapplicable in missile-borne bistatic Synthetic Aperture Radar (SAR) imaging with curved track due to the existing of three-axis velocity and acceleration. In addition, due to the existing of space-variant motion error introduced by acceleration, the traditional 2-D sub-block compensation method will cause the discontinuities between the image sub-blocks, thus affecting the subsequent image matching application. In view of these problems, this paper proposes a Back-Filtering PFA algorithm (BFPFA) which is based on the Improved Generalized Bistatic Equivalent Range Model (IGBERM). Constructing the combined compensating filter of space-variant phase error and geometric distortion, as well as reverse mapping interpolation, can realize the combined compensation of motion error, wavefront bending and geometric distortion in the process of oblique conversion, and obtain the SAR distance map without distortion, which is more conducive to the subsequent image matching applications. Finally, the simulations validate the effectiveness of the proposed algorithm.
2018, 40(11): 2645-2650.
doi: 10.11999/JEIT171203
Abstract:
The micro Synthetic Aperture Radar (SAR) system based on the traditional GaAs and GaN devices is not conducive to the monolithic integration, and the development bottleneck of volume, power consumption, weight and cost is becoming increasingly apparent, which is impossible to meet the needs of the miniaturized and ubiquitous unmanned platforms in the future. A new scheme for the design of a fully coherent Frequency Modulated Continuous Wave (FMCW) SAR with high resolution is proposed. The design method of high pulse phase stability and high isolation is studied and realized. The prototype of micro SAR is developed based on silicon chip and experimentally demonstrated. The micro SAR operates at K band, producing a signal bandwidth of wider than 2 GHz, enabling a range resolution of 7.5 cm. The system has made remarkable progress in terms of size, weight, power consumption and lay technical foundation for the monolithic integration of micro SAR system in a silicon chip.
The micro Synthetic Aperture Radar (SAR) system based on the traditional GaAs and GaN devices is not conducive to the monolithic integration, and the development bottleneck of volume, power consumption, weight and cost is becoming increasingly apparent, which is impossible to meet the needs of the miniaturized and ubiquitous unmanned platforms in the future. A new scheme for the design of a fully coherent Frequency Modulated Continuous Wave (FMCW) SAR with high resolution is proposed. The design method of high pulse phase stability and high isolation is studied and realized. The prototype of micro SAR is developed based on silicon chip and experimentally demonstrated. The micro SAR operates at K band, producing a signal bandwidth of wider than 2 GHz, enabling a range resolution of 7.5 cm. The system has made remarkable progress in terms of size, weight, power consumption and lay technical foundation for the monolithic integration of micro SAR system in a silicon chip.
2018, 40(11): 2651-2658.
doi: 10.11999/JEIT180062
Abstract:
Clutter of airborne bistatic radar is related to configuration and has serious range dependence characteristic, therefore the clutter ridge is complex and variable, and few Independent and Identically Distributed (IID) samples exist. As the result, the traditional Space-Time Adaptive Processing (STAP) has a degraded suppression performance for airborne bistatic radar clutter. Based on the sparsity of airborne radar clutter in the angle-Doppler domain and the advantages of Sparse Bayesian Learning (SBL) in sparse signal reconstruction, SBL algorithm is applied to the more complex airborne bistatic radar with both transmitter and receiver moving. The method can estimate the Clutter Covariance Matrix (CCM) of the unit under test with very few training samples, then perform space-time adaptive processing. Since the method does not need independent and identically distributed samples, it has better performance of clutter suppression in the airborne bistatic radar with both transmitter and receiver moving. Simulation results verify the effectiveness of the algorithm.
Clutter of airborne bistatic radar is related to configuration and has serious range dependence characteristic, therefore the clutter ridge is complex and variable, and few Independent and Identically Distributed (IID) samples exist. As the result, the traditional Space-Time Adaptive Processing (STAP) has a degraded suppression performance for airborne bistatic radar clutter. Based on the sparsity of airborne radar clutter in the angle-Doppler domain and the advantages of Sparse Bayesian Learning (SBL) in sparse signal reconstruction, SBL algorithm is applied to the more complex airborne bistatic radar with both transmitter and receiver moving. The method can estimate the Clutter Covariance Matrix (CCM) of the unit under test with very few training samples, then perform space-time adaptive processing. Since the method does not need independent and identically distributed samples, it has better performance of clutter suppression in the airborne bistatic radar with both transmitter and receiver moving. Simulation results verify the effectiveness of the algorithm.
2018, 40(11): 2659-2666.
doi: 10.11999/JEIT180076
Abstract:
Digital BeamForming (DBF) in elevation plays a crucial role for spaceborne Multiple Elevation Beam (MEB) SAR realizing the High-Resolution Wide-Swath (HRWS) imaging mode. However, due to the influence of satellite attitude error, the deviation of the DBF receiving beam direction always arises in such system. This leads to ghost targets appearing in the SAR image, when mapping the scenes (such as the seaport areas) with strong scatterers. To address the problem, a matrix pencil method based DBF processing approach in elevation is presented. Firstly, according to the given threshold, the peak position of the strong scatterer is found from the range-compressed signals. Then, the direction of arrival angle of the strong scatterer is estimated using the matrix pencil method. Finally, based on the imaging geometry model, the DBF weighting vector is adjusted to ensure the receiving beam to precisely point to the signal sources. Thereby, the interferences of ghost targets in SAR image can be eliminated effectively. The theoretical analysis is derived in detail, then it is validated by simulation experiments.
Digital BeamForming (DBF) in elevation plays a crucial role for spaceborne Multiple Elevation Beam (MEB) SAR realizing the High-Resolution Wide-Swath (HRWS) imaging mode. However, due to the influence of satellite attitude error, the deviation of the DBF receiving beam direction always arises in such system. This leads to ghost targets appearing in the SAR image, when mapping the scenes (such as the seaport areas) with strong scatterers. To address the problem, a matrix pencil method based DBF processing approach in elevation is presented. Firstly, according to the given threshold, the peak position of the strong scatterer is found from the range-compressed signals. Then, the direction of arrival angle of the strong scatterer is estimated using the matrix pencil method. Finally, based on the imaging geometry model, the DBF weighting vector is adjusted to ensure the receiving beam to precisely point to the signal sources. Thereby, the interferences of ghost targets in SAR image can be eliminated effectively. The theoretical analysis is derived in detail, then it is validated by simulation experiments.
2018, 40(11): 2667-2675.
doi: 10.11999/JEIT180101
Abstract:
Spread Doppler Clutter (SDC) caused by multi-mode propagation restrains the detection performance of Over-The-Horizon Radar (OTHR) for low detectable targets, such as slow ships. To solve this problem, a bi-iterative Minimum Variance Distortionless Response (MVDR) beamformer is proposed to suppress multi-mode SDC for MIMO OTHR system. As it is difficult to obtain the signal-free training data and enough sample support in MIMO-OTHR with time-staggered linear frequency modulated continuous wave or slow time phase-coded waveforms, the block matrix is used for data preprocessing to reduce the effect of expected signal component in the training data, then multi-mode SDC could be suppressed by the LN-variate MVDR beamformer which is restored through bi-iterative calculation with an L-variate transmit and an N-variate receive beamformer. This algorithm improves the convergence of MVDR beamformer, while reducing the computational load and the requirement of sample support. Theoretical analysis and simulation experiment are presented to verify the effectiveness of this algorithm.
Spread Doppler Clutter (SDC) caused by multi-mode propagation restrains the detection performance of Over-The-Horizon Radar (OTHR) for low detectable targets, such as slow ships. To solve this problem, a bi-iterative Minimum Variance Distortionless Response (MVDR) beamformer is proposed to suppress multi-mode SDC for MIMO OTHR system. As it is difficult to obtain the signal-free training data and enough sample support in MIMO-OTHR with time-staggered linear frequency modulated continuous wave or slow time phase-coded waveforms, the block matrix is used for data preprocessing to reduce the effect of expected signal component in the training data, then multi-mode SDC could be suppressed by the LN-variate MVDR beamformer which is restored through bi-iterative calculation with an L-variate transmit and an N-variate receive beamformer. This algorithm improves the convergence of MVDR beamformer, while reducing the computational load and the requirement of sample support. Theoretical analysis and simulation experiment are presented to verify the effectiveness of this algorithm.
2018, 40(11): 2676-2683.
doi: 10.11999/JEIT180200
Abstract:
Detection of stationary little targets in heavy ground clutter is the key problem facing the millimeter wave airport runway Foreign Object Debris (FOD) detection radar. This paper proposes a hierarchical FOD detection algorithm based on power spectrum feature extraction and Support Vector Domain Description (SVDD) classifier. The clutter map Constant False Alarm Rate (CFAR) detection algorithm is first utilized to suppress the complex background clutter. In order to solve the high false alarm problem after the clutter suppression, the power spectrum features are extracted to transform the radar returns into the feature domain where the FOD and false alarm are more distinguishable. Finally, the one-class SVDD classifier is utilized to categorize the FOD and false alarm into different kinds so as to reduce the false alarm rate. Experimental results based on measured data show that the proposed method can achieve good detection performance.
Detection of stationary little targets in heavy ground clutter is the key problem facing the millimeter wave airport runway Foreign Object Debris (FOD) detection radar. This paper proposes a hierarchical FOD detection algorithm based on power spectrum feature extraction and Support Vector Domain Description (SVDD) classifier. The clutter map Constant False Alarm Rate (CFAR) detection algorithm is first utilized to suppress the complex background clutter. In order to solve the high false alarm problem after the clutter suppression, the power spectrum features are extracted to transform the radar returns into the feature domain where the FOD and false alarm are more distinguishable. Finally, the one-class SVDD classifier is utilized to categorize the FOD and false alarm into different kinds so as to reduce the false alarm rate. Experimental results based on measured data show that the proposed method can achieve good detection performance.
2018, 40(11): 2684-2690.
doi: 10.11999/JEIT180117
Abstract:
Aircraft detection is a hot issue in the field of remote sensing image analysis. There exist many problems in current detection methods, such as complex detection procedure, low accuracy in complex background and dense aircraft area. To solve these problems, an end-to-end aircraft detection method named MDSSD is proposed in this paper. Based on Single Shot multibox Detector (SSD), a Densely connected convolutional Network (DenseNet) is used as the base network to extract features for its powerful ability in feature extraction, then an extra sub-network consisting of several feature layers is appended to detect and locate aircrafts. In order to locate aircrafts of various scales more accurately, a series of aspect ratios of default boxes are set to better match aircraft shapes and combine predictions deduced from feature maps of different layers. The method is more brief and efficient than methods that require object proposals, because it eliminates proposal generation completely and encapsulates all computation in a single network. Experiments demonstrate that this approach achieves better performance in many complex scenes.
Aircraft detection is a hot issue in the field of remote sensing image analysis. There exist many problems in current detection methods, such as complex detection procedure, low accuracy in complex background and dense aircraft area. To solve these problems, an end-to-end aircraft detection method named MDSSD is proposed in this paper. Based on Single Shot multibox Detector (SSD), a Densely connected convolutional Network (DenseNet) is used as the base network to extract features for its powerful ability in feature extraction, then an extra sub-network consisting of several feature layers is appended to detect and locate aircrafts. In order to locate aircrafts of various scales more accurately, a series of aspect ratios of default boxes are set to better match aircraft shapes and combine predictions deduced from feature maps of different layers. The method is more brief and efficient than methods that require object proposals, because it eliminates proposal generation completely and encapsulates all computation in a single network. Experiments demonstrate that this approach achieves better performance in many complex scenes.
2018, 40(11): 2691-2697.
doi: 10.11999/JEIT180002
Abstract:
Estimation of Direction Of Arrival (DOA) with scanned beams of single rotational antenna is meaningful. To obtain precise estimation with low computation burden, a closed-form estimator is proposed based on estimating the mode component. Firstly, the problem can be transformed into the estimation of mode component when antenna pattern is expressed with a formula of exponential sums, thus DOA can be induced from each mode. Considering the estimation error, a multi-mode estimator with its theoretical error is derived. Non-ideal observing conditions result in an ill-determined problem for the estimation of mode component. A modified method is proposed by reconstructing the antenna pattern. By calculating cross-correlation of the observed amplitude trains with the antenna pattern samples, a coarse estimation of DOA is obtained to determine the angle range under the matched reconstruction. Then, ill-determined problem can be avoided if the converted mode component is calculated with the new pattern. Both theoretical and simulation results demonstrate that the proposed method can obtain high precise estimation with low computation cost, and the proposed matched reconstruction approach extends the adaptability of the method.
Estimation of Direction Of Arrival (DOA) with scanned beams of single rotational antenna is meaningful. To obtain precise estimation with low computation burden, a closed-form estimator is proposed based on estimating the mode component. Firstly, the problem can be transformed into the estimation of mode component when antenna pattern is expressed with a formula of exponential sums, thus DOA can be induced from each mode. Considering the estimation error, a multi-mode estimator with its theoretical error is derived. Non-ideal observing conditions result in an ill-determined problem for the estimation of mode component. A modified method is proposed by reconstructing the antenna pattern. By calculating cross-correlation of the observed amplitude trains with the antenna pattern samples, a coarse estimation of DOA is obtained to determine the angle range under the matched reconstruction. Then, ill-determined problem can be avoided if the converted mode component is calculated with the new pattern. Both theoretical and simulation results demonstrate that the proposed method can obtain high precise estimation with low computation cost, and the proposed matched reconstruction approach extends the adaptability of the method.
2018, 40(11): 2698-2704.
doi: 10.11999/JEIT180074
Abstract:
Due to the distortions of the broadcasted satellite signals and the inconsistencies of parameter settings for different receivers, the single difference or double difference of pseudo-ranges between two receivers are different for two pair of different receivers. Bias inconsistencies will lead to adverse effects for pseudo-range-based positioning applications. Pseudo-range biases can also hinder carrier-phase ambiguity resolution. However, fewer articles deal with pseudo-range biases for BeiDou navigation satellite System (BDS). In order to mitigate the impact of biases on BDS to the greatest extent, the generation mechanisms and characteristics of pseudo-range biases are studied in detail firstly. Then based on this, experimental verification methods are designed using Haoping Radio Observatory (HRO) of Chinese Academy of Sciences to observe BDS signals. Pseudo-range biases of all visible BDS satellites are measured and evaluated with high accuracy, using the 40 meters dish antenna and modern equipment of HRO. Finally, some important parameters of BDS receivers, such as the correlator spacing and front-end bandwidth, are suggested to mitigate the ranging errors and positioning errors result from pseudo-range biases. The achievements of this paper can provide a worthy reference for GNSS signal designers, GNSS monitoring and assessment and GNSS receiver designers.
Due to the distortions of the broadcasted satellite signals and the inconsistencies of parameter settings for different receivers, the single difference or double difference of pseudo-ranges between two receivers are different for two pair of different receivers. Bias inconsistencies will lead to adverse effects for pseudo-range-based positioning applications. Pseudo-range biases can also hinder carrier-phase ambiguity resolution. However, fewer articles deal with pseudo-range biases for BeiDou navigation satellite System (BDS). In order to mitigate the impact of biases on BDS to the greatest extent, the generation mechanisms and characteristics of pseudo-range biases are studied in detail firstly. Then based on this, experimental verification methods are designed using Haoping Radio Observatory (HRO) of Chinese Academy of Sciences to observe BDS signals. Pseudo-range biases of all visible BDS satellites are measured and evaluated with high accuracy, using the 40 meters dish antenna and modern equipment of HRO. Finally, some important parameters of BDS receivers, such as the correlator spacing and front-end bandwidth, are suggested to mitigate the ranging errors and positioning errors result from pseudo-range biases. The achievements of this paper can provide a worthy reference for GNSS signal designers, GNSS monitoring and assessment and GNSS receiver designers.
2018, 40(11): 2705-2711.
doi: 10.11999/JEIT180094
Abstract:
A novel broadband circularly polarized monopole antenna is proposed by microstrip feed line. The antenna is composed of C-shaped patch and an improved ground plane with the overall size of 25×25×1 mm3. The impedance bandwidth and axial ratio bandwidth of the antenna can be effectively widened by cutting the corner on the C-shaped patch and adding triangular stubs on the ground plane. The design procedure of the antenna is given, and the working mechanism of the circularly polarized antenna is analyzed from the surface current distributions. Besides, the antenna is fabricated and measured. Simulated and measured results show that the antenna has ultra wide impedance bandwidth and axial ratio bandwidth. The operating bandwidth of the antenna is 4.35～12 GHz (relative bandwidth 93.6%), and the 3 dB axial ratio bandwidth is 4.15～11.8 GHz (relative bandwidth 95.9%). At the same time, the radiation performance and gain characteristics of the antenna are measured and the measured results are in good agreement with the simulated results, which proves the effectiveness of the antenna. The antenna can be applied to Ultra-WideBand (UWB) wireless communication systems and satellite communication systems.
A novel broadband circularly polarized monopole antenna is proposed by microstrip feed line. The antenna is composed of C-shaped patch and an improved ground plane with the overall size of 25×25×1 mm3. The impedance bandwidth and axial ratio bandwidth of the antenna can be effectively widened by cutting the corner on the C-shaped patch and adding triangular stubs on the ground plane. The design procedure of the antenna is given, and the working mechanism of the circularly polarized antenna is analyzed from the surface current distributions. Besides, the antenna is fabricated and measured. Simulated and measured results show that the antenna has ultra wide impedance bandwidth and axial ratio bandwidth. The operating bandwidth of the antenna is 4.35～12 GHz (relative bandwidth 93.6%), and the 3 dB axial ratio bandwidth is 4.15～11.8 GHz (relative bandwidth 95.9%). At the same time, the radiation performance and gain characteristics of the antenna are measured and the measured results are in good agreement with the simulated results, which proves the effectiveness of the antenna. The antenna can be applied to Ultra-WideBand (UWB) wireless communication systems and satellite communication systems.
2018, 40(11): 2712-2719.
doi: 10.11999/JEIT180025
Abstract:
Focusing on the problem of reducing the large computation cost of traditional antenna design methods, a new surrogate model based on Back Propagation Neural Networks (BPNN) is constructed. In order to solve the problem of easily falling into local optimum in BPNN, a PSO-BPNN surrogate model is developed by improving initial structural parameters of neural networks and applied to fast multi-objective optimization design of multi-parameter antenna structures. The design results show that the proposed PSO-BPNN outperforms other existing antenna surrogate models in terms of prediction accuracy and prediction speed. The proposed method is of value in dealing with complex antenna designs with high-dimensional parameter space.
Focusing on the problem of reducing the large computation cost of traditional antenna design methods, a new surrogate model based on Back Propagation Neural Networks (BPNN) is constructed. In order to solve the problem of easily falling into local optimum in BPNN, a PSO-BPNN surrogate model is developed by improving initial structural parameters of neural networks and applied to fast multi-objective optimization design of multi-parameter antenna structures. The design results show that the proposed PSO-BPNN outperforms other existing antenna surrogate models in terms of prediction accuracy and prediction speed. The proposed method is of value in dealing with complex antenna designs with high-dimensional parameter space.
2018, 40(11): 2720-2727.
doi: 10.11999/JEIT180060
Abstract:
Continuous Phase Frequency Shift Keying (CPFSK) is widely adopted as a standard by the telemetry community. The Multi-Symbol Detection (MSD) technique can increase channel gain for the CPFSK telemetry system. Therefore, the timing synchronization method for CPFSK signal needs to adapt to the scenario with lower SNR. According to existing timing synchronization methods’ poor performance in low SNR, a novel timing synchronization method for CPFSK signal based on MSD is proposed, which is suitable to variable rate. The simulation results show that, when Eb/N0 is 0 dB and symbol rate is 2 Mbps, the proposed method achieves 2 dB more channel gain than the single symbol likelihood decision method, and has similar performance to the early-late gate code synchronization method with reduced hardware resource by 60%. Finally, the validity of the numerical simulation and resource evaluation is verified by principle prototype realization.
Continuous Phase Frequency Shift Keying (CPFSK) is widely adopted as a standard by the telemetry community. The Multi-Symbol Detection (MSD) technique can increase channel gain for the CPFSK telemetry system. Therefore, the timing synchronization method for CPFSK signal needs to adapt to the scenario with lower SNR. According to existing timing synchronization methods’ poor performance in low SNR, a novel timing synchronization method for CPFSK signal based on MSD is proposed, which is suitable to variable rate. The simulation results show that, when Eb/N0 is 0 dB and symbol rate is 2 Mbps, the proposed method achieves 2 dB more channel gain than the single symbol likelihood decision method, and has similar performance to the early-late gate code synchronization method with reduced hardware resource by 60%. Finally, the validity of the numerical simulation and resource evaluation is verified by principle prototype realization.
Accurate Acquisition of High Order Double Binary Offset Carrier Signals for High Dynamic Environment
2018, 40(11): 2728-2735.
doi: 10.11999/JEIT180087
Abstract:
For the problem of without accurate acquisition of Double Binary Offset Carrier (DBOC) modulated signal for high dynamic environment, a method which is based on Partial Matched Filtering (PMF) - Fast Fourier Transform (FFT) is proposed. According to the problem of low detection performance caused by the related loss and scallop loss, a new improved acquisition scheme is proposed. Firstly, the Discrete Polynomial phase Transform (DPT) is used to remove the high order dynamic term of the received signal, and then the PMF-FFT algorithm is redesigned for the DBOC signal. Finally, the spectrum correction method is used to correct the maximum power spectrum after FFT. Simulation results show that, under the same conditions, the proposed scheme improves the detection probability by about 2 dB, and shortens effectively the acquisition time.
For the problem of without accurate acquisition of Double Binary Offset Carrier (DBOC) modulated signal for high dynamic environment, a method which is based on Partial Matched Filtering (PMF) - Fast Fourier Transform (FFT) is proposed. According to the problem of low detection performance caused by the related loss and scallop loss, a new improved acquisition scheme is proposed. Firstly, the Discrete Polynomial phase Transform (DPT) is used to remove the high order dynamic term of the received signal, and then the PMF-FFT algorithm is redesigned for the DBOC signal. Finally, the spectrum correction method is used to correct the maximum power spectrum after FFT. Simulation results show that, under the same conditions, the proposed scheme improves the detection probability by about 2 dB, and shortens effectively the acquisition time.
2018, 40(11): 2736-2743.
doi: 10.11999/JEIT180027
Abstract:
Mobile Edge Computing (MEC) draws much attention in the next generation of mobile networks with high bandwidth and low latency by enabling the IT and cloud computation capacity at the Radio Access Network (RAN). Matching problem between requesting nodes and servicing nodes is studied when a vehicle wants to offload tasks, a MEC-based offloading framework in vehicular networks is proposed, Vehicle can either offload task to MEC sever as V2I link or neighboring vehicle as V2V link. Taking into account the limited and heterogeneous resources, and the diversity of tasks, offloading framework is established as combination auction model, and a multi-round sequential combination auction mechanism is proposed, which consists of Analytic Hierarchy Process (AHP) ranking, task bidding and winners decision. Simulation results show that the proposed mechanism can maximize the efficiency of service nodes while increasing the efficiency of requesting vehicles under the constraints of the delay and the capacity.
Mobile Edge Computing (MEC) draws much attention in the next generation of mobile networks with high bandwidth and low latency by enabling the IT and cloud computation capacity at the Radio Access Network (RAN). Matching problem between requesting nodes and servicing nodes is studied when a vehicle wants to offload tasks, a MEC-based offloading framework in vehicular networks is proposed, Vehicle can either offload task to MEC sever as V2I link or neighboring vehicle as V2V link. Taking into account the limited and heterogeneous resources, and the diversity of tasks, offloading framework is established as combination auction model, and a multi-round sequential combination auction mechanism is proposed, which consists of Analytic Hierarchy Process (AHP) ranking, task bidding and winners decision. Simulation results show that the proposed mechanism can maximize the efficiency of service nodes while increasing the efficiency of requesting vehicles under the constraints of the delay and the capacity.
2018, 40(11): 2744-2751.
doi: 10.11999/JEIT180110
Abstract:
A multiuser Differential Chaos Shift Keying (DCSK) communication system based on Hilbert transform is proposed (HMU-DCSK), to solve the problem of low transmission rate of DCSK. Under the condition of fixed-order Walsh codes, the set of orthogonally-based signals is doubled by the Hilbert transform and the carrier signals assigned to each user are guaranteed to be orthogonal. The Bit-Error-Rate (BER) formula in Rayleigh fading channel is derived and numerous simulations are conducted. The simulation results show that the transmission rate of HMU-DCSK system is twice that of traditional multiuser DCSK system under the same N value, meanwhile, the BER performance of HMU-DCSK system is obviously better than the traditional multi-user DCSK system under the same transmission rate.
A multiuser Differential Chaos Shift Keying (DCSK) communication system based on Hilbert transform is proposed (HMU-DCSK), to solve the problem of low transmission rate of DCSK. Under the condition of fixed-order Walsh codes, the set of orthogonally-based signals is doubled by the Hilbert transform and the carrier signals assigned to each user are guaranteed to be orthogonal. The Bit-Error-Rate (BER) formula in Rayleigh fading channel is derived and numerous simulations are conducted. The simulation results show that the transmission rate of HMU-DCSK system is twice that of traditional multiuser DCSK system under the same N value, meanwhile, the BER performance of HMU-DCSK system is obviously better than the traditional multi-user DCSK system under the same transmission rate.
2018, 40(11): 2752-2757.
doi: 10.11999/JEIT180197
Abstract:
By using the characteristic of matrix eigenvalues, this paper proposes a new secret sharing scheme without trusted center. The scheme does not require a trusted center, and each participant provides the same secret share (column vector) and generates its own secret share in the black box, thus avoiding the authority deception of the trusted center. Reversible matrix P consisting of column vectors provided by all participants,and diagonal matrix \begin{document}${Λ}$\end{document} ![]()
![]()
generate a matrix M . Then, the orthogonalized unit eigenvectors of the matrix M is distributed to each participant as a subkey. Because the eigenvalues corresponding to the participants in the same set are the same, this scheme can effectively prevent malicious fraud among members. Analysis results show that the program is feasible and safe.
By using the characteristic of matrix eigenvalues, this paper proposes a new secret sharing scheme without trusted center. The scheme does not require a trusted center, and each participant provides the same secret share (column vector) and generates its own secret share in the black box, thus avoiding the authority deception of the trusted center. Reversible matrix P consisting of column vectors provided by all participants,and diagonal matrix
2018, 40(11): 2758-2764.
doi: 10.11999/JEIT180130
Abstract:
The performance of trajectory based user identification is poor since the existing methods ignore the order feature of location sequence. To solve this problem, a Cross Domain Trajectory matching algorithm based on Paragraph2vec (CDTraj2vec) is proposed. Firstly, the user trajectory is transformed to the grid representation which is easy to handle. The PV-DM model in the Paragraph2vec algorithm is utilized for extracting order feature of location sequence in trajectory. Then the original user trajectories are divided by a certain time size and distance scale to construct a training sample suitable for training PV-DM model. The PV-DM model is trained by different types of training samples, and the vector representation of the user trajectories is obtained. Finally, the matching of the trajectory is determined by the user trajectory vector. Experimental results on BrightKite shows that the F-measure is improved by 2%～4% compared with the existing frequency based and distance based algorithm. The proposed algorithm can effectively extract the order feature of location sequence, and realize the trajectory based user identification across social networks.
The performance of trajectory based user identification is poor since the existing methods ignore the order feature of location sequence. To solve this problem, a Cross Domain Trajectory matching algorithm based on Paragraph2vec (CDTraj2vec) is proposed. Firstly, the user trajectory is transformed to the grid representation which is easy to handle. The PV-DM model in the Paragraph2vec algorithm is utilized for extracting order feature of location sequence in trajectory. Then the original user trajectories are divided by a certain time size and distance scale to construct a training sample suitable for training PV-DM model. The PV-DM model is trained by different types of training samples, and the vector representation of the user trajectories is obtained. Finally, the matching of the trajectory is determined by the user trajectory vector. Experimental results on BrightKite shows that the F-measure is improved by 2%～4% compared with the existing frequency based and distance based algorithm. The proposed algorithm can effectively extract the order feature of location sequence, and realize the trajectory based user identification across social networks.
2018, 40(11): 2765-2771.
doi: 10.11999/JEIT180108
Abstract:
Considering the difficulty of neighbor discovery in underwater acoustic communication networks, a neighbor discovery mechanism is presented based on directional transmission and reception. In this mechanism, the nodes only send and receive signals directionally, which can avoid the hidden terminal problem caused by asymmetric gain and increase the network coverage. Time is divided into neighbor discovery time slot and listening & reply time slot. In neighbor discovery time slot, the node sends the HELLO signal, and then waits to receive the REPLY signal sent by its neighbor node. In listening & reply time slot, the node listens the channel for the HELLO signal sent by the source node, then replies REPLY signal to the source node. The node can discover its neighbor through HELLO/REPLY two-way handshake based on competition and direct & indirect discovery, which can overcome the " deaf” nodes problem and improve the efficiency of neighbor discovery. Compared with the randomized two-way neighbor discovery mechanism, simulation tests show that the proposed mechanism has the shorter average discovery latency and the higher average discovery ratio in various network density and number of antenna sectors.
Considering the difficulty of neighbor discovery in underwater acoustic communication networks, a neighbor discovery mechanism is presented based on directional transmission and reception. In this mechanism, the nodes only send and receive signals directionally, which can avoid the hidden terminal problem caused by asymmetric gain and increase the network coverage. Time is divided into neighbor discovery time slot and listening & reply time slot. In neighbor discovery time slot, the node sends the HELLO signal, and then waits to receive the REPLY signal sent by its neighbor node. In listening & reply time slot, the node listens the channel for the HELLO signal sent by the source node, then replies REPLY signal to the source node. The node can discover its neighbor through HELLO/REPLY two-way handshake based on competition and direct & indirect discovery, which can overcome the " deaf” nodes problem and improve the efficiency of neighbor discovery. Compared with the randomized two-way neighbor discovery mechanism, simulation tests show that the proposed mechanism has the shorter average discovery latency and the higher average discovery ratio in various network density and number of antenna sectors.
2018, 40(11): 2772-2778.
doi: 10.11999/JEIT180042
Abstract:
Protocol Oblivious Forwarding (POF) supports the arbitrary protocol processing, enhancing the programmability of Software Defined Networking (SDN). In order to improve the forwarding performance, a flow caching method is proposed. To parse the packet in advance, absolute positions of matching fields are obtained by identifying the dependency of matching and actions. To guarantee the acceleration effect of flow caching, flow tables are selected according to their matching types and number of entries. In addition, the single-flow table cache and multi-flow table cache are compared and an adaptive switching strategy is proposed based on the actual situation of network traffic. The POFSwitch is extended to implement the proposed method and it is validated under the real rules and backbone traces. The switch packet forwarding rate is increased by 220% after applying flow caching. Flow caching can provide higher forwarding performance for programmable data planes.
Protocol Oblivious Forwarding (POF) supports the arbitrary protocol processing, enhancing the programmability of Software Defined Networking (SDN). In order to improve the forwarding performance, a flow caching method is proposed. To parse the packet in advance, absolute positions of matching fields are obtained by identifying the dependency of matching and actions. To guarantee the acceleration effect of flow caching, flow tables are selected according to their matching types and number of entries. In addition, the single-flow table cache and multi-flow table cache are compared and an adaptive switching strategy is proposed based on the actual situation of network traffic. The POFSwitch is extended to implement the proposed method and it is validated under the real rules and backbone traces. The switch packet forwarding rate is increased by 220% after applying flow caching. Flow caching can provide higher forwarding performance for programmable data planes.
2018, 40(11): 2779-2786.
doi: 10.11999/JEIT180026
Abstract:
Marine magnetic anomaly detection is one of the basic means of marine scientific observation, exploration of undersea resources, national defense and security. However, the complexity of the magnetic field noise increases the difficulty of the magnetic detection. It is of great significance to study various magnetic field noise mechanisms and suppression methods for the improvement of measurement accuracy. In this paper, the wave magnetic field model under general and infinite depth conditions is used to estimate the noise induced by sea waves respectively. The wave and geomagnetic noise in the magnetic anomaly signal is filtered out by the combination of spectral subtraction and wavelet. In order to verify the validity of the algorithm, the ocean magnetic field in a sea area of South China Sea in August 2015 is observed. The results show that this method can filter out most of the wave and geomagnetic field noise. The wave distribution in the frequency range of 0.4～0.8 Hz is obviously reduced, the waveform in the time domain is greatly improved, the magnetic anomaly signal of the target is highlighted. Signal to noise ratio can be increased by nearly 11 dB. The proposed method has the advantages of low computational complexity, strong real-time performance and easy implementation, which can provide an effective measure for noise suppression of marine magnetic anomaly detection.
Marine magnetic anomaly detection is one of the basic means of marine scientific observation, exploration of undersea resources, national defense and security. However, the complexity of the magnetic field noise increases the difficulty of the magnetic detection. It is of great significance to study various magnetic field noise mechanisms and suppression methods for the improvement of measurement accuracy. In this paper, the wave magnetic field model under general and infinite depth conditions is used to estimate the noise induced by sea waves respectively. The wave and geomagnetic noise in the magnetic anomaly signal is filtered out by the combination of spectral subtraction and wavelet. In order to verify the validity of the algorithm, the ocean magnetic field in a sea area of South China Sea in August 2015 is observed. The results show that this method can filter out most of the wave and geomagnetic field noise. The wave distribution in the frequency range of 0.4～0.8 Hz is obviously reduced, the waveform in the time domain is greatly improved, the magnetic anomaly signal of the target is highlighted. Signal to noise ratio can be increased by nearly 11 dB. The proposed method has the advantages of low computational complexity, strong real-time performance and easy implementation, which can provide an effective measure for noise suppression of marine magnetic anomaly detection.
2018, 40(11): 2787-2794.
doi: 10.11999/JEIT180045
Abstract:
Continuous monitoring of IntraOcular Pressure (IOP) plays an important role in the diagnosis and treatment of the glaucoma. Existing IOP sensors have some problems, such as low sensitivities, high central resonant frequencies and difficult fabrication. In order to solve the aforementioned problems, this paper presents a wireless, passive and non-invasive IOP sensor based on MEMS technology. The sensor contains five stacked layers, where Parylene, copper and PDMS are adopted as the functional materials within two flexible substrate layers, two electrode layers, and a dielectric layer, respectively. The electrode layers and the dielectric layer consist of two inductors and two capacitors to form a resonant circuit in C-L-C-L series. In the term of fabrication, a MEMS planar process followed by thermally shaping is proposed to fit curved surfaces of the eyeballs, and then this design scheme can effectively solve such issues as the difficulty in making the sensor and so on. Experimental results show that the central resonant frequency is decreased to 40 MHz, relative sensitivity is quantified as 1028.57 ppm/kPa, and resolution reached up to 50 Pa (0.375 mmHg). This study can be used for long-term, continuous monitoring of IOP.
Continuous monitoring of IntraOcular Pressure (IOP) plays an important role in the diagnosis and treatment of the glaucoma. Existing IOP sensors have some problems, such as low sensitivities, high central resonant frequencies and difficult fabrication. In order to solve the aforementioned problems, this paper presents a wireless, passive and non-invasive IOP sensor based on MEMS technology. The sensor contains five stacked layers, where Parylene, copper and PDMS are adopted as the functional materials within two flexible substrate layers, two electrode layers, and a dielectric layer, respectively. The electrode layers and the dielectric layer consist of two inductors and two capacitors to form a resonant circuit in C-L-C-L series. In the term of fabrication, a MEMS planar process followed by thermally shaping is proposed to fit curved surfaces of the eyeballs, and then this design scheme can effectively solve such issues as the difficulty in making the sensor and so on. Experimental results show that the central resonant frequency is decreased to 40 MHz, relative sensitivity is quantified as 1028.57 ppm/kPa, and resolution reached up to 50 Pa (0.375 mmHg). This study can be used for long-term, continuous monitoring of IOP.
