Editor's Note
Guest Editor
Article List
2021, 19(4):569-572.
DOI: 10.11805/TKYDA2021184
Abstract:
A passive location method to long baseline antenna array is proposed. In this method, an array antenna is embedded into a long baseline system with tri-antenna firstly. According to the principle that the angle between the target and the long baseline is implied in the phase difference information, the angle changing rate can be estimated from the phase difference of the angle between the target and the long baseline. The actual direction of the target can be measured by the array antenna. The position parameters of the target are obtained thereafter. The method can be applied to locate the target by using only a single pulse. The simulation results verify the validity of the proposed method.
A passive location method to long baseline antenna array is proposed. In this method, an array antenna is embedded into a long baseline system with tri-antenna firstly. According to the principle that the angle between the target and the long baseline is implied in the phase difference information, the angle changing rate can be estimated from the phase difference of the angle between the target and the long baseline. The actual direction of the target can be measured by the array antenna. The position parameters of the target are obtained thereafter. The method can be applied to locate the target by using only a single pulse. The simulation results verify the validity of the proposed method.
2021, 19(4):589-595.
DOI: 10.11805/TKYDA2021146
Abstract:
At present, artificial intelligence-based methods have been able to achieve good results in radar emitter recognition task. However, with the development of electronic information technology, there will be more and more unknown emitters whose characteristic distribution and categories are unknown. In the absence of prior knowledge, it is difficult to fully train the artificial intelligence model, which makes most of the existing methods unable to well complete the recognition of unknown radar emitters. This paper proposes a big electromagnetic data solution that can be used for the recognition of unknown radar emitters, and then focuses on the Flink-based fast comparison retrieval and recognition algorithm for unknown radar emitters. Finally, a comparative experiment proves the effectiveness of the proposed method, and its recognition accuracy can reach 87.2%. When the parallelism is set to 6, the entire Mutual Information- K-Nearest Neighbor(MI-KNN) parallelization algorithm takes only 4.7 s.
At present, artificial intelligence-based methods have been able to achieve good results in radar emitter recognition task. However, with the development of electronic information technology, there will be more and more unknown emitters whose characteristic distribution and categories are unknown. In the absence of prior knowledge, it is difficult to fully train the artificial intelligence model, which makes most of the existing methods unable to well complete the recognition of unknown radar emitters. This paper proposes a big electromagnetic data solution that can be used for the recognition of unknown radar emitters, and then focuses on the Flink-based fast comparison retrieval and recognition algorithm for unknown radar emitters. Finally, a comparative experiment proves the effectiveness of the proposed method, and its recognition accuracy can reach 87.2%. When the parallelism is set to 6, the entire Mutual Information- K-Nearest Neighbor(MI-KNN) parallelization algorithm takes only 4.7 s.
3 Individual identification method of communication radiation source based on power spectral density
2021, 19(4):596-602.
DOI: 10.11805/TKYDA2021140
Abstract:
An individual identification method of communication radiation sources based on Power Spectral Density(PSD) fingerprint characteristics and intelligent classifier is proposed in order to prevent the occurrence of problems such as device cloning, replay attacks and user identity impersonation, and to accurately identify and authenticate Internet of Things(IoT) objects. First, the radio frequency baseband signal is collected by receiver, and the in-phase signal is collected. Then the steady-state signal segment is intercepted through variance trajectory detection, and data normalization processing on the steady-state signal segment is performed; the PSD of the steady-state signal segment is calculated after data normalization processing to obtain a feature vector, and the feature vector is used as the radio frequency fingerprint of the transmitter. Finally, an intelligent classifier is adopted to identify the radio frequency fingerprint to complete the individual identification of the communication radiation source. The experimental test to identify eight wireless data transmission radio E90-DTU devices and 100 WiFi network card devices of the same manufacturer, the same type and the same batch shows that the proposed method can obtain good recognition accuracy when applied in Line-Of-Sight(LOS) scenarios, mixed scenes of LOS and Non-line-Of-Sight(NOS) scenarios, low signal-to-noise ratio scenes, and scenarios with a large number of IoT devices, etc.
An individual identification method of communication radiation sources based on Power Spectral Density(PSD) fingerprint characteristics and intelligent classifier is proposed in order to prevent the occurrence of problems such as device cloning, replay attacks and user identity impersonation, and to accurately identify and authenticate Internet of Things(IoT) objects. First, the radio frequency baseband signal is collected by receiver, and the in-phase signal is collected. Then the steady-state signal segment is intercepted through variance trajectory detection, and data normalization processing on the steady-state signal segment is performed; the PSD of the steady-state signal segment is calculated after data normalization processing to obtain a feature vector, and the feature vector is used as the radio frequency fingerprint of the transmitter. Finally, an intelligent classifier is adopted to identify the radio frequency fingerprint to complete the individual identification of the communication radiation source. The experimental test to identify eight wireless data transmission radio E90-DTU devices and 100 WiFi network card devices of the same manufacturer, the same type and the same batch shows that the proposed method can obtain good recognition accuracy when applied in Line-Of-Sight(LOS) scenarios, mixed scenes of LOS and Non-line-Of-Sight(NOS) scenarios, low signal-to-noise ratio scenes, and scenarios with a large number of IoT devices, etc.
2021, 19(4):603-616.
DOI: 10.11805/TKYDA2021139
Abstract:
Radio Frequency Fingerprint(RFF) originates from the differences in transmitter circuit design and the manufacturing tolerance of the hardware circuit in the production process. It is an emerging equipment identification and authentication technology. Modeling the generation mechanism of RFF is the basis for its in-depth research. Based on a general Zero Intermediate Frequency(ZIF) digital communication transmitter architecture, the influence of each component in the transmitter on RFF is analyzed, and the corresponding RFF time-domain baseband model is established as well. In addition, several important time-domain parameter tolerances of communication standards are summarized. The maximum Root Mean Square Error Vector Magnitudes(RMS EVMs) of the two typical modulation methods, Quadrature Phase Shift Keying(QPSK) and 16 Quadrature Amplitude Modulation(16-QAM), are mainly studied under the LTE standard. Finally, through theoretical derivation and Matlab simulation, the upper and lower bounds of Direct Current(DC) offset, In-Phase/Quadrature(I/Q) gain imbalance, I/Q quadrature offset error, I/Q filter offset, oscillator phase noise, and power amplifier nonlinearity parameters are given. The changes of the constellation diagram under the critical conditions of various RFF parameters are also analyzed, which provides reasonable parameter guidance for the future research of RFF extraction and identification.
Radio Frequency Fingerprint(RFF) originates from the differences in transmitter circuit design and the manufacturing tolerance of the hardware circuit in the production process. It is an emerging equipment identification and authentication technology. Modeling the generation mechanism of RFF is the basis for its in-depth research. Based on a general Zero Intermediate Frequency(ZIF) digital communication transmitter architecture, the influence of each component in the transmitter on RFF is analyzed, and the corresponding RFF time-domain baseband model is established as well. In addition, several important time-domain parameter tolerances of communication standards are summarized. The maximum Root Mean Square Error Vector Magnitudes(RMS EVMs) of the two typical modulation methods, Quadrature Phase Shift Keying(QPSK) and 16 Quadrature Amplitude Modulation(16-QAM), are mainly studied under the LTE standard. Finally, through theoretical derivation and Matlab simulation, the upper and lower bounds of Direct Current(DC) offset, In-Phase/Quadrature(I/Q) gain imbalance, I/Q quadrature offset error, I/Q filter offset, oscillator phase noise, and power amplifier nonlinearity parameters are given. The changes of the constellation diagram under the critical conditions of various RFF parameters are also analyzed, which provides reasonable parameter guidance for the future research of RFF extraction and identification.
2021, 19(4):617-622.
DOI: 10.11805/TKYDA2021122
Abstract:
An intelligent modulation recognition method based on the Convolutional Neural Network(CNN) and two-dimensional Red-Green-Blue(RGB) cyclic spectrum images is proposed in order to improve the modulation recognition accuracy and reduce the computational complexity. The cyclic spectrum can be employed to identify the modulation type. The three-dimensional cyclic spectra are converted to two-dimensional RGB cyclic spectra to reduce the computational complexity, which are then taken to build the data set. Moreover, a CNN based modulation classifier with low computational complexity is proposed. Simulation results show that the proposed intelligent modulation recognition algorithm can achieve higher classification accuracy with lower computational complexity.
An intelligent modulation recognition method based on the Convolutional Neural Network(CNN) and two-dimensional Red-Green-Blue(RGB) cyclic spectrum images is proposed in order to improve the modulation recognition accuracy and reduce the computational complexity. The cyclic spectrum can be employed to identify the modulation type. The three-dimensional cyclic spectra are converted to two-dimensional RGB cyclic spectra to reduce the computational complexity, which are then taken to build the data set. Moreover, a CNN based modulation classifier with low computational complexity is proposed. Simulation results show that the proposed intelligent modulation recognition algorithm can achieve higher classification accuracy with lower computational complexity.
2021, 19(4):623-627.
DOI: 10.11805/TKYDA2021084
Abstract:
Power Spectral Density(PSD) prediction is an important part of spectrum management. Due to the high complexity, nonlinearity and uncertainty of the PSD, it is difficult for a single prediction model to ensure the accuracy and efficiency of the prediction. In order to overcome the disadvantages of a single prediction method, a hybrid machine learning model is proposed to combine a Self-Organizing Map(SOM) network with a Regression Tree(RT) to predict the PSD of the signal. First, the method uses a self-organizing map network to cluster the original sample sets with similar manual features. Then, a RT is constructed for each cluster to predict the PSD. Finally, the data of RWTH from Aachen University are adopted for experiments. The root mean square error of the prediction result is 0.824 higher than that of the existing method, which proves that the hybrid model has higher prediction accuracy and better generalization ability.
Power Spectral Density(PSD) prediction is an important part of spectrum management. Due to the high complexity, nonlinearity and uncertainty of the PSD, it is difficult for a single prediction model to ensure the accuracy and efficiency of the prediction. In order to overcome the disadvantages of a single prediction method, a hybrid machine learning model is proposed to combine a Self-Organizing Map(SOM) network with a Regression Tree(RT) to predict the PSD of the signal. First, the method uses a self-organizing map network to cluster the original sample sets with similar manual features. Then, a RT is constructed for each cluster to predict the PSD. Finally, the data of RWTH from Aachen University are adopted for experiments. The root mean square error of the prediction result is 0.824 higher than that of the existing method, which proves that the hybrid model has higher prediction accuracy and better generalization ability.
2021, 19(4):556-561.
DOI: 10.11805/TKYDA2021214
Abstract:
Transfer learning technology can use experience information to assist current tasks. It has been widely used in the fields of computer vision and speech recognition, whereas it has not made obvious achievements in the electromagnetic field. The electromagnetic environment changes quickly, and the performance of the source data or the classifier model in the new environment will be significantly degraded. Retraining not only requires a lot of data but also takes time and effort. Transfer learning technology is very related to the task of electromagnetic target recognition. Based on the measured electromagnetic target data set, this paper explores several applications of transfer learning in solving the problem of small samples of electromagnetic targets, including the transfer of similar targets and the transfer of heterogeneous targets. Experimental results show that by migrating the pre-training model to the target domain small sample recognition task, when the target domain is a similar source and there are only 20 labeled samples, the verification accuracy is increased by 25% compared with the non-transfer model and the training time is greatly shortened; when the target domain is a heterogeneous source, the training time can be less than 1/5 that of the source domain while ensuring the recognition accuracy.
Transfer learning technology can use experience information to assist current tasks. It has been widely used in the fields of computer vision and speech recognition, whereas it has not made obvious achievements in the electromagnetic field. The electromagnetic environment changes quickly, and the performance of the source data or the classifier model in the new environment will be significantly degraded. Retraining not only requires a lot of data but also takes time and effort. Transfer learning technology is very related to the task of electromagnetic target recognition. Based on the measured electromagnetic target data set, this paper explores several applications of transfer learning in solving the problem of small samples of electromagnetic targets, including the transfer of similar targets and the transfer of heterogeneous targets. Experimental results show that by migrating the pre-training model to the target domain small sample recognition task, when the target domain is a similar source and there are only 20 labeled samples, the verification accuracy is increased by 25% compared with the non-transfer model and the training time is greatly shortened; when the target domain is a heterogeneous source, the training time can be less than 1/5 that of the source domain while ensuring the recognition accuracy.
2021, 19(4):581-588.
DOI: 10.11805/TKYDA2021150
Abstract:
The realization of abnormal detection and pattern discovery of electromagnetic data is of great value to the judgment and early warning of abnormal behaviors of electromagnetic targets. Different types of electromagnetic data usually exist in the form of time series, with the characteristic of imbalance between normal data and abnormal data. To address these issues，a time series anomaly detection method is proposed based on the spatial-temporal joint attention mechanism. The channel attention mechanism and spatial attention mechanism are combined to enhance the feature representation of the abnormal part of time series data. Experimental results show that the proposed detection algorithm can effectively deal with the difficulty of data imbalance and has strong robustness.
The realization of abnormal detection and pattern discovery of electromagnetic data is of great value to the judgment and early warning of abnormal behaviors of electromagnetic targets. Different types of electromagnetic data usually exist in the form of time series, with the characteristic of imbalance between normal data and abnormal data. To address these issues，a time series anomaly detection method is proposed based on the spatial-temporal joint attention mechanism. The channel attention mechanism and spatial attention mechanism are combined to enhance the feature representation of the abnormal part of time series data. Experimental results show that the proposed detection algorithm can effectively deal with the difficulty of data imbalance and has strong robustness.
2021, 19(4):635-641.
DOI: 10.11805/TKYDA2021080
Abstract:
This paper proposes a multi-scale Convolutional Neural Network(CNN) online pre- earthquake electromagnetic anomaly detection model which is applied in noisy environment. Based on the powerful feature extraction ability of CNN, cooperating with the characteristics of long-term and short-term ground-space electromagnetic spectrum, the pre-earthquake electromagnetic anomaly detection is performed in multi-dimensional and multi-perspective. At the same time, the adaptive Variational Mode Decomposition(VMD) noise reduction method is introduced to extract the effective information in the observation signal. Combined with online learning strategy, the continuous learning of possible changes of pre-earthquake electromagnetic anomaly mode is realized. The simulation results show that the multi-scale model can maintain high accuracy under low Signal-to-Noise Ratio(SNR), and the online learning strategy can effectively reduce the model update time, which proves the effectiveness of the model.
This paper proposes a multi-scale Convolutional Neural Network(CNN) online pre- earthquake electromagnetic anomaly detection model which is applied in noisy environment. Based on the powerful feature extraction ability of CNN, cooperating with the characteristics of long-term and short-term ground-space electromagnetic spectrum, the pre-earthquake electromagnetic anomaly detection is performed in multi-dimensional and multi-perspective. At the same time, the adaptive Variational Mode Decomposition(VMD) noise reduction method is introduced to extract the effective information in the observation signal. Combined with online learning strategy, the continuous learning of possible changes of pre-earthquake electromagnetic anomaly mode is realized. The simulation results show that the multi-scale model can maintain high accuracy under low Signal-to-Noise Ratio(SNR), and the online learning strategy can effectively reduce the model update time, which proves the effectiveness of the model.
2021, 19(4):562-568.
DOI: 10.11805/TKYDA2021195
Abstract:
In complex communication environment, the connection between the characteristics of different signals is seldom considered in modulation recognition. A Convolutional Neural Network(CNN) is built to extract the characteristics of the time-frequency images of signals. Time-frequency transform is employed to process the one-dimensional signal into images, and image features are extracted through CNN. In order to improve the classification and recognition accuracy of the algorithm under low SNR, the texture features are also extracted from the images, and they are fused with the features extracted from the CNN. The simulation results show that the Time–Frequency Convolution Neural Network(TF–CNN) and TF–Resnet framework can achieve signal automatic modulation recognition and classification.
In complex communication environment, the connection between the characteristics of different signals is seldom considered in modulation recognition. A Convolutional Neural Network(CNN) is built to extract the characteristics of the time-frequency images of signals. Time-frequency transform is employed to process the one-dimensional signal into images, and image features are extracted through CNN. In order to improve the classification and recognition accuracy of the algorithm under low SNR, the texture features are also extracted from the images, and they are fused with the features extracted from the CNN. The simulation results show that the Time–Frequency Convolution Neural Network(TF–CNN) and TF–Resnet framework can achieve signal automatic modulation recognition and classification.
2021, 19(4):549-555.
DOI: 10.11805/TKYDA2021156
Abstract:
Due to the increasingly complex electromagnetic environment, the electromagnetic space situation has the characteristics of integrity, dynamics, relevance, visibility, mass, multi-dimensionality and so on. The study on the electromagnetic space situation provides an important theoretical support for improving the comprehensive performance of the frequency system, realizing the spectrum sharing of the mobile communication system and ensuring the spectrum security of major security activities, which has gradually become one of the most dynamic research directions in the spectrum field. The related technologies of electromagnetic space situation research at home and abroad are highlighted in this paper, and the representative work is systematically introduced. The importance and development status of electromagnetic space situation research are summarized, and the challenges in this field in the future are put forward.
Due to the increasingly complex electromagnetic environment, the electromagnetic space situation has the characteristics of integrity, dynamics, relevance, visibility, mass, multi-dimensionality and so on. The study on the electromagnetic space situation provides an important theoretical support for improving the comprehensive performance of the frequency system, realizing the spectrum sharing of the mobile communication system and ensuring the spectrum security of major security activities, which has gradually become one of the most dynamic research directions in the spectrum field. The related technologies of electromagnetic space situation research at home and abroad are highlighted in this paper, and the representative work is systematically introduced. The importance and development status of electromagnetic space situation research are summarized, and the challenges in this field in the future are put forward.
2021, 19(4):628-634.
DOI: 10.11805/TKYDA2021046
Abstract:
Most of the current direct positioning methods are mainly for narrow-band signals. A wide/narrow band signal Direct Position Determination(DPD) method based on distributed Unmanned Aerial Vehicle(UAV) platform is proposed. Firstly, the received data of multi-UAV platforms are synthesized in frequency domain, and a cost function directly related to the source location is established based on multiple frequency points. Secondly, the monitoring area is gridded to determine the location of the target source. Finally, the multi-UAV movement monitoring is carried out, and the monitoring area is continuously reduced. The final positioning result is obtained by clustering analysis of multiple positioning results. Simulation results show that the localization performance of this method is obviously better than that of traditional localization methods. The processing results of the measured data show that the proposed method has better positioning performance than the improved Time Difference of Arrival(TDOA) positioning method.
Most of the current direct positioning methods are mainly for narrow-band signals. A wide/narrow band signal Direct Position Determination(DPD) method based on distributed Unmanned Aerial Vehicle(UAV) platform is proposed. Firstly, the received data of multi-UAV platforms are synthesized in frequency domain, and a cost function directly related to the source location is established based on multiple frequency points. Secondly, the monitoring area is gridded to determine the location of the target source. Finally, the multi-UAV movement monitoring is carried out, and the monitoring area is continuously reduced. The final positioning result is obtained by clustering analysis of multiple positioning results. Simulation results show that the localization performance of this method is obviously better than that of traditional localization methods. The processing results of the measured data show that the proposed method has better positioning performance than the improved Time Difference of Arrival(TDOA) positioning method.
2021, 19(4):573-580.
DOI: 10.11805/TKYDA2021172
Abstract:
Multiple heterogeneous spectrum users require different Quality of Service(QoS) in cognitive radio networks. A dynamic spectrum allocation method is proposed based on multi-agent reinforcement learning. In order to improve the satisfaction of spectrum users, the proposed method is evaluated by the Quality of Experience(QoE) of spectrum users instead of QoS. Multiple virtual agents are established to simulate spectrum users to learn interactively with environment in a cooperative way, and the optimal spectrum allocation can be obtained by integrating their learning and spectrum decision results. Simulation results show that the proposed method can obtain higher QoE performance of secondary users than those methods based on the traditional reinforcement learning. The probability of collision between spectrum users also can be reduced in the proposed method without any information about the usage rules of primary users and dynamic characteristics of channels.
Multiple heterogeneous spectrum users require different Quality of Service(QoS) in cognitive radio networks. A dynamic spectrum allocation method is proposed based on multi-agent reinforcement learning. In order to improve the satisfaction of spectrum users, the proposed method is evaluated by the Quality of Experience(QoE) of spectrum users instead of QoS. Multiple virtual agents are established to simulate spectrum users to learn interactively with environment in a cooperative way, and the optimal spectrum allocation can be obtained by integrating their learning and spectrum decision results. Simulation results show that the proposed method can obtain higher QoE performance of secondary users than those methods based on the traditional reinforcement learning. The probability of collision between spectrum users also can be reduced in the proposed method without any information about the usage rules of primary users and dynamic characteristics of channels.
2022, 20(1):34-39.
DOI: 10.11805/TKYDA2021151
Abstract:
The electromagnetic space of large-scale Low Earth Orbit(LEO) satellite constellation system is complex and difficult to observe. The link characteristics of LEO satellite constellation are studied. Firstly, taking Starlink and OneWeb constellations as the research objects, according to the constellation parameters of LEO satellites, Equivalent Isotropically Radiated Power(EIRP) values are obtained and visualized. Then the data of LEO electromagnetic satellite are analyzed to obtain the numerical relationship between attenuation characteristics and time, frequency, etc. The inter-satellite link interference and its temporal distribution characteristics are calculated. The characteristic values of relative interference time are obtained, as well as the attenuation and time-frequency multi-dimensional characteristics of inter-satellite data. The time characteristics of interference in different scenarios are analyzed. The inter-satellite link interference of large-scale LEO satellite constellation system is studied from multiple dimensions and verified by simulation. Finally, it is proved that there is interference in the inter-satellite links among large-scale LEO satellite constellation systems, and the higher the frequency is, the more obvious the interference is.
The electromagnetic space of large-scale Low Earth Orbit(LEO) satellite constellation system is complex and difficult to observe. The link characteristics of LEO satellite constellation are studied. Firstly, taking Starlink and OneWeb constellations as the research objects, according to the constellation parameters of LEO satellites, Equivalent Isotropically Radiated Power(EIRP) values are obtained and visualized. Then the data of LEO electromagnetic satellite are analyzed to obtain the numerical relationship between attenuation characteristics and time, frequency, etc. The inter-satellite link interference and its temporal distribution characteristics are calculated. The characteristic values of relative interference time are obtained, as well as the attenuation and time-frequency multi-dimensional characteristics of inter-satellite data. The time characteristics of interference in different scenarios are analyzed. The inter-satellite link interference of large-scale LEO satellite constellation system is studied from multiple dimensions and verified by simulation. Finally, it is proved that there is interference in the inter-satellite links among large-scale LEO satellite constellation systems, and the higher the frequency is, the more obvious the interference is.
2022, 20(1):22-28.
DOI: 10.11805/TKYDA2021153
Abstract:
Based on tensor splitting technique, a decentralized computing method of neural network model for electromagnetic object detection is introduced. In this method, different tensor splitting techniques are selected according to different hidden layers, and the weights are distributed to multiple distributed nodes losslessly. The simulation results on Raspberry PI show that this method can decompose and deploy the centralized detection model losslessly, and ensure the same accuracy as the original model. And when the original model is too heavy to be loaded into memory for calculation, this method can still complete the calculation properly.
Based on tensor splitting technique, a decentralized computing method of neural network model for electromagnetic object detection is introduced. In this method, different tensor splitting techniques are selected according to different hidden layers, and the weights are distributed to multiple distributed nodes losslessly. The simulation results on Raspberry PI show that this method can decompose and deploy the centralized detection model losslessly, and ensure the same accuracy as the original model. And when the original model is too heavy to be loaded into memory for calculation, this method can still complete the calculation properly.
2022, 20(1):29-33,39.
DOI: 10.11805/TKYDA2021217
Abstract:
In recent years, many high-quality datasets have supported the rapid development of deep learning in the field of computer vision, speech and natural language processing. Nevertheless, there is still a lack of high-quality datasets in the field of electromagnetic signal recognition. In order to promote in-depth learning in the application of electromagnetic signal recognition, a large-scale real electromagnetic signal dataset is established based on Automatic Dependent Surveillance-Broadcast (ADS-B). An automatic data collection and labeling system is designed to automatically capture ADS-B electromagnetic signals in open and real scenes. A high quality ADS-B signal dataset is established by data cleaning and sorting of ADS-B signals. The performance of in-depth learning models using datasets is studied, and the models are evaluated comprehensively under different signal-to-noise ratios, sampling rates and number of samples. The data set provides a valuable benchmark for relevant researchers.
In recent years, many high-quality datasets have supported the rapid development of deep learning in the field of computer vision, speech and natural language processing. Nevertheless, there is still a lack of high-quality datasets in the field of electromagnetic signal recognition. In order to promote in-depth learning in the application of electromagnetic signal recognition, a large-scale real electromagnetic signal dataset is established based on Automatic Dependent Surveillance-Broadcast (ADS-B). An automatic data collection and labeling system is designed to automatically capture ADS-B electromagnetic signals in open and real scenes. A high quality ADS-B signal dataset is established by data cleaning and sorting of ADS-B signals. The performance of in-depth learning models using datasets is studied, and the models are evaluated comprehensively under different signal-to-noise ratios, sampling rates and number of samples. The data set provides a valuable benchmark for relevant researchers.
2022, 20(1):40-46.
DOI: 10.11805/TKYDA2021152
Abstract:
Frequency-hopping signal shows good performance in anti-interference. Accurately identifying the modulation methods of frequency-hopping signals can provide strong support for military information warfare such as judging the attributes of enemy and enemy targets and interfering with enemy signals. Nevertheless, there is still a big gap in the modulation recognition of frequency hopping signals at home and abroad. A frequency-hopping signal modulation recognition method based on time-frequency features is proposed. Through Smoothed Pseudo Wigner-Ville Distribution(SPWVD) time-frequency transformation, time-frequency images of frequency-hopping signals of different modulation types are obtained, and the time-frequency images are sent to a Convolutional Neural Network(CNN) for feature extraction and classification recognition. Simulation experiments prove that the proposed CNN model has achieved better recognition results under low Signal-to-Noise Ratios(SNRs).
Frequency-hopping signal shows good performance in anti-interference. Accurately identifying the modulation methods of frequency-hopping signals can provide strong support for military information warfare such as judging the attributes of enemy and enemy targets and interfering with enemy signals. Nevertheless, there is still a big gap in the modulation recognition of frequency hopping signals at home and abroad. A frequency-hopping signal modulation recognition method based on time-frequency features is proposed. Through Smoothed Pseudo Wigner-Ville Distribution(SPWVD) time-frequency transformation, time-frequency images of frequency-hopping signals of different modulation types are obtained, and the time-frequency images are sent to a Convolutional Neural Network(CNN) for feature extraction and classification recognition. Simulation experiments prove that the proposed CNN model has achieved better recognition results under low Signal-to-Noise Ratios(SNRs).
2022, 20(1):16-21,28.
DOI: 10.11805/TKYDA2021189
Abstract:
With the proliferation of frequency-using devices and the advent of the era of big data, spectrum management and control are faced with challenges of effectiveness and accuracy. Modulation classification technology is the foundation and key part of spectrum management and control. Therefore, the effectiveness of modulation classification technology in big data scenario is very important. This paper considers not only the validity of the classification model under the background of big data, but also the dynamics of noise in the complex electromagnetic environment. A big dataset containing different signals under different Mixed Signal-to-Noise Ratios(MSNR) is constructed, and the big data is utilized to drive the Deep Learning model, and the classification results are finally obtained. The proposed method can realize modulation classification by training just one model, which avoids the redundancy of model training in previous algorithms. The simulation results demonstrate the effectiveness and reliability of the proposed method.
With the proliferation of frequency-using devices and the advent of the era of big data, spectrum management and control are faced with challenges of effectiveness and accuracy. Modulation classification technology is the foundation and key part of spectrum management and control. Therefore, the effectiveness of modulation classification technology in big data scenario is very important. This paper considers not only the validity of the classification model under the background of big data, but also the dynamics of noise in the complex electromagnetic environment. A big dataset containing different signals under different Mixed Signal-to-Noise Ratios(MSNR) is constructed, and the big data is utilized to drive the Deep Learning model, and the classification results are finally obtained. The proposed method can realize modulation classification by training just one model, which avoids the redundancy of model training in previous algorithms. The simulation results demonstrate the effectiveness and reliability of the proposed method.
2022, 20(1):53-57,89.
DOI: 10.11805/TKYDA2021155
Abstract:
The spectrum situation prediction of non-cooperative wireless network in the complex electromagnetic environment is investigated. Based on machine learning theory, the three-dimensional characteristics of time, space, and frequency of collected spectrum situation data are extracted; the inherent correlations in the three-dimensional characteristics are fully data mined; and the spectrum prediction frameworks are built to predict frequency adjustment behavior of non-cooperative communication nodes. The results show that the single-step or multi-step prediction for the frequency can be performed on the frequency adjustment for future moments by exploiting the spectrum prediction frameworks as long as sufficient spectrum situation data can be intercepted when the frequency adjustment exists in the communication process of non-cooperative wireless networks. Therefore, the possible frequency used in the future for the target system can be accurately locked in. This work can provide key technical support for the subsequent communication tracking and interference tasks.
The spectrum situation prediction of non-cooperative wireless network in the complex electromagnetic environment is investigated. Based on machine learning theory, the three-dimensional characteristics of time, space, and frequency of collected spectrum situation data are extracted; the inherent correlations in the three-dimensional characteristics are fully data mined; and the spectrum prediction frameworks are built to predict frequency adjustment behavior of non-cooperative communication nodes. The results show that the single-step or multi-step prediction for the frequency can be performed on the frequency adjustment for future moments by exploiting the spectrum prediction frameworks as long as sufficient spectrum situation data can be intercepted when the frequency adjustment exists in the communication process of non-cooperative wireless networks. Therefore, the possible frequency used in the future for the target system can be accurately locked in. This work can provide key technical support for the subsequent communication tracking and interference tasks.
2022, 20(1):47-52.
DOI: 10.11805/TKYDA2021165
Abstract:
Base station location optimization is a research hotspot in mobile communication. A good base station location scheme can not only save resources, but also improve users' communication experience. However, the base station layout is often faced with a complex problem of multi-parameter, multi-constraint and nonlinearity, which is difficult to be solved by traditional optimization methods. In this paper, an intelligent base station layout method based on big data is proposed. Firstly, the radio wave propagation model based on deep learning is built according to the measured big data of electromagnetic environment, which makes the propagation model more accurate. Then, the spatial adaptive learning method is utilized to construct the base station location optimization model on the basis of the propagation model. By selecting the base station placement points having poor performance with a small probability in each iteration process, the algorithm can avoid falling into local optimality. The experimental simulation results show that the proposed base station layout method has fast convergence speed, wide coverage rate and good user communication experience.
Base station location optimization is a research hotspot in mobile communication. A good base station location scheme can not only save resources, but also improve users' communication experience. However, the base station layout is often faced with a complex problem of multi-parameter, multi-constraint and nonlinearity, which is difficult to be solved by traditional optimization methods. In this paper, an intelligent base station layout method based on big data is proposed. Firstly, the radio wave propagation model based on deep learning is built according to the measured big data of electromagnetic environment, which makes the propagation model more accurate. Then, the spatial adaptive learning method is utilized to construct the base station location optimization model on the basis of the propagation model. By selecting the base station placement points having poor performance with a small probability in each iteration process, the algorithm can avoid falling into local optimality. The experimental simulation results show that the proposed base station layout method has fast convergence speed, wide coverage rate and good user communication experience.
2022, 20(1):8-15.
DOI: 10.11805/TKYDA2021168
Abstract:
With the development of urban communication technology and the increase of frequency equipment, the electromagnetic environment becomes more and more complex. Fully understanding the characteristics of spectrum resource utilization in the past is the key to improve the efficiency of spectrum management. A complete process about detailed data quality analysis for big data in complex and diverse electromagnetic environment is proposed, in order to explore the characteristics of spectrum utilization more comprehensively. The spectrum correlation for different channels in the same service, and for different channels in different services, is performed. Attribute construction is carried out for big data of electromagnetic environment, including the attributes of frequency dimension occupancy and time dimension occupancy. The multi-dimensional Gaussian mixture model in the field of image processing is introduced to remove the background noise of the electromagnetic signal and extract the electromagnetic signal, which can lay the foundation for the subsequent information mining and association analysis.
With the development of urban communication technology and the increase of frequency equipment, the electromagnetic environment becomes more and more complex. Fully understanding the characteristics of spectrum resource utilization in the past is the key to improve the efficiency of spectrum management. A complete process about detailed data quality analysis for big data in complex and diverse electromagnetic environment is proposed, in order to explore the characteristics of spectrum utilization more comprehensively. The spectrum correlation for different channels in the same service, and for different channels in different services, is performed. Attribute construction is carried out for big data of electromagnetic environment, including the attributes of frequency dimension occupancy and time dimension occupancy. The multi-dimensional Gaussian mixture model in the field of image processing is introduced to remove the background noise of the electromagnetic signal and extract the electromagnetic signal, which can lay the foundation for the subsequent information mining and association analysis.
2022, 20(1):1-7.
DOI: 10.11805/TKYDA2021145
Abstract:
The Internet of Vehicles(IoV) is a research hotspot of the fifth generation(5G) mobile communication network. Cellular Vehicle To Everything(C-V2X) is an Internet of Vehicles solution based on cellular network technology and is an important part of ultra-Reliable and Low Latency Communication(uRLLC) in 5G network. The realization of the Internet of Vehicles technology is of great significance to modern transportation. This paper offers a systematic survey of existing research achievements of the domestic and foreign researchers in recent years. Firstly, a brief description of the definition of the Internet of Vehicles is given, and the standard research progress of C-V2X is summarized. Next, the centralized and distributed resource scheduling methods under LTE-V2X and NR-V2X are described respectively, and the existing research methods are classified. Finally, a perspective of the future work in this research area is discussed.
The Internet of Vehicles(IoV) is a research hotspot of the fifth generation(5G) mobile communication network. Cellular Vehicle To Everything(C-V2X) is an Internet of Vehicles solution based on cellular network technology and is an important part of ultra-Reliable and Low Latency Communication(uRLLC) in 5G network. The realization of the Internet of Vehicles technology is of great significance to modern transportation. This paper offers a systematic survey of existing research achievements of the domestic and foreign researchers in recent years. Firstly, a brief description of the definition of the Internet of Vehicles is given, and the standard research progress of C-V2X is summarized. Next, the centralized and distributed resource scheduling methods under LTE-V2X and NR-V2X are described respectively, and the existing research methods are classified. Finally, a perspective of the future work in this research area is discussed.
.
Abstract:
Advanced Encryption Standard (AES) is the current encryption standard adopted by U.S. government and plays an important role in cryptograph systems. Meanwhile, Radar signal transmission requires reliability, high throughput and high security, which rise the demands of adapting AES as a stream cipher. Hereby, by redesigning the CTR mode, AES as a stream cipher is achieved. The architecture folding and algorithm reuse improve the throughput and keep the resource utilization low. On-the-fly key scheduling, pure logic S-Box, distributed share memory and pipeline processing are achieved in small FPGAs. The real-life test shows the design satisfies the requirements of Radar signal encryption.
Advanced Encryption Standard (AES) is the current encryption standard adopted by U.S. government and plays an important role in cryptograph systems. Meanwhile, Radar signal transmission requires reliability, high throughput and high security, which rise the demands of adapting AES as a stream cipher. Hereby, by redesigning the CTR mode, AES as a stream cipher is achieved. The architecture folding and algorithm reuse improve the throughput and keep the resource utilization low. On-the-fly key scheduling, pure logic S-Box, distributed share memory and pipeline processing are achieved in small FPGAs. The real-life test shows the design satisfies the requirements of Radar signal encryption.
.
Abstract:
The hybrid waveform combined LFM and pseudo-random binary phase code (LFM-PRBC) is a new type of low probability of intercept signal, which is widely adopted in radars and small detectors. It is significant to estimate the parameters of the hybrid waveform. This paper develops the modified Cramer-Rao Lower Bound (MCRLB) for the modulation parameters of this signal with Gaussian noise. The modulation parameters considered are signal amplitude, carrier frequency, modulation slope and code width, and the nuisance parameters are codes and initial phase. Conclusions indicate that MCRLB is independent with nuisance parameters and MCRLB is linear with the SNR in logarithmic form when the vector of estimated parameters is determined.
The hybrid waveform combined LFM and pseudo-random binary phase code (LFM-PRBC) is a new type of low probability of intercept signal, which is widely adopted in radars and small detectors. It is significant to estimate the parameters of the hybrid waveform. This paper develops the modified Cramer-Rao Lower Bound (MCRLB) for the modulation parameters of this signal with Gaussian noise. The modulation parameters considered are signal amplitude, carrier frequency, modulation slope and code width, and the nuisance parameters are codes and initial phase. Conclusions indicate that MCRLB is independent with nuisance parameters and MCRLB is linear with the SNR in logarithmic form when the vector of estimated parameters is determined.
2010, 8(6):625-632.
Abstract:
In this paper the efficient numerical methods for EM scattering from low observation targets have been introduced. These methods based on MLFMA and included some further improvements, among which the JMCFIE for surface integral equation of homogeneous dielectric scattering, the multi-TDS boundary conditions for volume integral equation of multi-layered dielectric scattering and their applications have been discussed in detail. Finally, the code, called as Parallelized Universal Electromagnetic Scattering Tool (P-UEST) and developed by our group for modeling of scattering from the stealth airplane, has been described. Some application examples have also been given in this paper.
In this paper the efficient numerical methods for EM scattering from low observation targets have been introduced. These methods based on MLFMA and included some further improvements, among which the JMCFIE for surface integral equation of homogeneous dielectric scattering, the multi-TDS boundary conditions for volume integral equation of multi-layered dielectric scattering and their applications have been discussed in detail. Finally, the code, called as Parallelized Universal Electromagnetic Scattering Tool (P-UEST) and developed by our group for modeling of scattering from the stealth airplane, has been described. Some application examples have also been given in this paper.
2011, 9(2):127-132.
Abstract:
Target decomposition is a key point for polarimetric Synthetic Aperture Radar(SAR)’s applications，where the four-component decomposition has been successfully applied to the analysis of urban areas and some other complex areas. However，the decomposed volume scattering component is relatively large when the four-component decomposition is applied. It is caused by the volume scattering model which is not good for characterizing totally random scatters. In order to accommodate the targets’ real physical scattering characteristics，a modified four-component decomposition based on a new volume scattering model is proposed in this paper. First，the deorientation process is applied to the coherency matrix. Then，with the new volume scattering model，a modified decomposition algorithm is proposed with a power constraint for avoiding the emergence of negative powers. Finally，the NASA/JPL AIRSAR L-band polarimetric data of San Francisco and the ALOS/PALSAR polarimetric data acquired over Beijing area are used for analysis；the experimental results demonstrate the effectiveness of the proposed algorithm.
Target decomposition is a key point for polarimetric Synthetic Aperture Radar(SAR)’s applications，where the four-component decomposition has been successfully applied to the analysis of urban areas and some other complex areas. However，the decomposed volume scattering component is relatively large when the four-component decomposition is applied. It is caused by the volume scattering model which is not good for characterizing totally random scatters. In order to accommodate the targets’ real physical scattering characteristics，a modified four-component decomposition based on a new volume scattering model is proposed in this paper. First，the deorientation process is applied to the coherency matrix. Then，with the new volume scattering model，a modified decomposition algorithm is proposed with a power constraint for avoiding the emergence of negative powers. Finally，the NASA/JPL AIRSAR L-band polarimetric data of San Francisco and the ALOS/PALSAR polarimetric data acquired over Beijing area are used for analysis；the experimental results demonstrate the effectiveness of the proposed algorithm.
2011, 9(1):1-6.
Abstract:
In the process of obtaining video，many factors lead to the degradation of the video quality，and lower the spatial resolution of video；on the other hand，the exposure time and the frame rate of camera limit the time resolution of video. Video super-resolution reconstruction is an effective way to improve the temporal resolution and the spatial resolution of video，which has aroused wide concern in computer vision and image processing fields. In this paper，the concept of video super-resolution is described and the development phases in recent years are reviewed. In addition，some of the key technologies are analyzed. Furthermore，the challenges in this field are pointed out. Finally，the application of video super- reconstruction technology is prospected.
In the process of obtaining video，many factors lead to the degradation of the video quality，and lower the spatial resolution of video；on the other hand，the exposure time and the frame rate of camera limit the time resolution of video. Video super-resolution reconstruction is an effective way to improve the temporal resolution and the spatial resolution of video，which has aroused wide concern in computer vision and image processing fields. In this paper，the concept of video super-resolution is described and the development phases in recent years are reviewed. In addition，some of the key technologies are analyzed. Furthermore，the challenges in this field are pointed out. Finally，the application of video super- reconstruction technology is prospected.
.
Abstract:
Focus on the illumination coherence problem of the three-dimensional virtual scene, the environment mapping technology was analysed to improve the reality. The light probe image is used to light the scene and the irradiance of the diffuse surface is approximated by using spherical harmonics. By considering the shadowed diffuse transfer simultaneously, the soft shadow of diffuse objects is generated successfully under dynamic, low-frequency lighting environments. It is showed that this algorithm approaches accurately with high quality.
Focus on the illumination coherence problem of the three-dimensional virtual scene, the environment mapping technology was analysed to improve the reality. The light probe image is used to light the scene and the irradiance of the diffuse surface is approximated by using spherical harmonics. By considering the shadowed diffuse transfer simultaneously, the soft shadow of diffuse objects is generated successfully under dynamic, low-frequency lighting environments. It is showed that this algorithm approaches accurately with high quality.
2022, 20(12):1257-1268.
DOI: 10.11805/TKYDA2021143
Abstract:
Benchmark datasets are crucial for many data-dependent scientific studies and technology applications. Academic and industry communities have closely collaborated to release abundant datasets in many fields. However, there is still a lack of high-quality benchmark datasets in some specific domains. This paper introduces two open-source benchmark datasets, namely, the Insider Threat Dataset (ITD-2018) and the Indoor Crowd Movement Trajectory Dataset(ICMTD-2019). The two datasets are produced by program-driven synthetic data generation methods and are presented with well-defined scenarios, carefully-designed behavior models, rich data patterns, and vivid storylines. The two datasets were used in the ChinaVis Data Challenge. This paper aims to promote the two datasets for the development of the research and technology in relevant domains.
Benchmark datasets are crucial for many data-dependent scientific studies and technology applications. Academic and industry communities have closely collaborated to release abundant datasets in many fields. However, there is still a lack of high-quality benchmark datasets in some specific domains. This paper introduces two open-source benchmark datasets, namely, the Insider Threat Dataset (ITD-2018) and the Indoor Crowd Movement Trajectory Dataset(ICMTD-2019). The two datasets are produced by program-driven synthetic data generation methods and are presented with well-defined scenarios, carefully-designed behavior models, rich data patterns, and vivid storylines. The two datasets were used in the ChinaVis Data Challenge. This paper aims to promote the two datasets for the development of the research and technology in relevant domains.
2022, 20(12):1285-1291.
DOI: 10.11805/TKYDA2021180
Abstract:
A processing process of open-set specific emitter identification is built in order to achieve accurate control of urban frequency equipment. The core lies in the effective interval filtering of fingerprint features and the open set recognition model based on the deep self-encoder. By visualizing deep network activation using Class Activation Mapping(Grad-CAM), the section of signal contributing more to neural network activation can be determined, and then interval filtering for the signal can be performed without losing too much fingerprint information. On the other hand, an open-set specific emitter identification model is established based on semi-supervised adversarial autoencoders, achieving effective monitoring and identification of unknown emitters that may occur in the spectrum. Experiments show that Grad-CAM can filter out the most advantageous part of the extracted signal fingerprint, and the proposed model can achieve high-precision open set recognition without degrading the closed set recognition rate.
A processing process of open-set specific emitter identification is built in order to achieve accurate control of urban frequency equipment. The core lies in the effective interval filtering of fingerprint features and the open set recognition model based on the deep self-encoder. By visualizing deep network activation using Class Activation Mapping(Grad-CAM), the section of signal contributing more to neural network activation can be determined, and then interval filtering for the signal can be performed without losing too much fingerprint information. On the other hand, an open-set specific emitter identification model is established based on semi-supervised adversarial autoencoders, achieving effective monitoring and identification of unknown emitters that may occur in the spectrum. Experiments show that Grad-CAM can filter out the most advantageous part of the extracted signal fingerprint, and the proposed model can achieve high-precision open set recognition without degrading the closed set recognition rate.
2022, 20(12):1298-1304.
DOI: 10.11805/TKYDA2021352
Abstract:
Aiming at the problems of long signal sequence and poor feature robustness in Feature Engineering in individual recognition, the technology based on deep neural network is studied. Drawing lessons from Convolutional, Long Short-Term Memory, Fully Connected Deep Neural Networks(CLDNN) in speech recognition, the local amplitude features of the signal are extracted through convolution neural network and the global time-domain features of the signal are extracted through long-term and short-term memory network. A fully connected network is utilized to map the feature to the device label. Under the line of sight channel, the data of eight Lora modulated wireless data transmission stations are collected, and the Gaussian white noise is added to the simulation test. The simulation shows that when the Signal-to-Noise Ratio is low(0 dB) , the accuracy of the model can reach nearly 95% under the signal sequence length of 2 048 points. In addition, this model needs fewer parameters compared with VGG16 model. The proposed model has a certain application prospect in the deployment of Internet of things devices.
Aiming at the problems of long signal sequence and poor feature robustness in Feature Engineering in individual recognition, the technology based on deep neural network is studied. Drawing lessons from Convolutional, Long Short-Term Memory, Fully Connected Deep Neural Networks(CLDNN) in speech recognition, the local amplitude features of the signal are extracted through convolution neural network and the global time-domain features of the signal are extracted through long-term and short-term memory network. A fully connected network is utilized to map the feature to the device label. Under the line of sight channel, the data of eight Lora modulated wireless data transmission stations are collected, and the Gaussian white noise is added to the simulation test. The simulation shows that when the Signal-to-Noise Ratio is low(0 dB) , the accuracy of the model can reach nearly 95% under the signal sequence length of 2 048 points. In addition, this model needs fewer parameters compared with VGG16 model. The proposed model has a certain application prospect in the deployment of Internet of things devices.
2022, 20(12):1305-1310.
DOI: 10.11805/TKYDA2021356
Abstract:
The radio frequency fingerprints are inherent features of the device hardware, and will not change with the transmitted signal, therefore they are often used in communication anti-spoofing. In this paper, the neural network is adopted to process the original signal samples obtained by the receiver, including I/Q sequence, amplitude/phase, binary image of constellation diagram and color density diagram of constellation diagram to achieve anti-deception effect. When the signal-to-interference and noise ratio is in the range of -30 dB to 30 dB, the signal recognition accuracy can reach up to 99.93%. Being different from the existing literature, the method can be adapted to the scenes with different signal-to-interference and noise ratios. This research shows that the proposed method is feasible to achieve anti-spoofing in a complex communication environment where spoofing signals and legal signals coexist.
The radio frequency fingerprints are inherent features of the device hardware, and will not change with the transmitted signal, therefore they are often used in communication anti-spoofing. In this paper, the neural network is adopted to process the original signal samples obtained by the receiver, including I/Q sequence, amplitude/phase, binary image of constellation diagram and color density diagram of constellation diagram to achieve anti-deception effect. When the signal-to-interference and noise ratio is in the range of -30 dB to 30 dB, the signal recognition accuracy can reach up to 99.93%. Being different from the existing literature, the method can be adapted to the scenes with different signal-to-interference and noise ratios. This research shows that the proposed method is feasible to achieve anti-spoofing in a complex communication environment where spoofing signals and legal signals coexist.
2022, 20(12):1318-1325.
DOI: 10.11805/TKYDA2022028
Abstract:
In order to solve the adaptive problem of generating jamming waveforms in response to enemy communication signals in Unmanned Aerial Vehicle(UAV) swarm communication confrontation, a cognitive jamming waveform design method is proposed based on Spectrally Modulated Spectrally Encoded(SMSE) model. By deploying six kinds of different waveform design parameters in the SMSE framework, the waveforms with a specific spectrum structure are synthesized in the frequency domain to generate cognitive jamming waveforms with corresponding functions to deal with communication signals under different parameters in communication countermeasures. Simulation experiments show that the theoretical model can generate the suppressive jamming waveforms: single-tone jamming, multi-tone jamming, wideband interference, and narrowband interference. It can also achieve modulation deceptive jamming for the modulation methods of BPSK, QPSK, 8PSK, and 16QAM signals. Compared with the theoretical curve, the experimental results verify the theoretical feasibility of SMSE model to generate cognitive interference waveform. SMSE model can realize the integration of suppressive jamming and spoofing jamming.
In order to solve the adaptive problem of generating jamming waveforms in response to enemy communication signals in Unmanned Aerial Vehicle(UAV) swarm communication confrontation, a cognitive jamming waveform design method is proposed based on Spectrally Modulated Spectrally Encoded(SMSE) model. By deploying six kinds of different waveform design parameters in the SMSE framework, the waveforms with a specific spectrum structure are synthesized in the frequency domain to generate cognitive jamming waveforms with corresponding functions to deal with communication signals under different parameters in communication countermeasures. Simulation experiments show that the theoretical model can generate the suppressive jamming waveforms: single-tone jamming, multi-tone jamming, wideband interference, and narrowband interference. It can also achieve modulation deceptive jamming for the modulation methods of BPSK, QPSK, 8PSK, and 16QAM signals. Compared with the theoretical curve, the experimental results verify the theoretical feasibility of SMSE model to generate cognitive interference waveform. SMSE model can realize the integration of suppressive jamming and spoofing jamming.
2022, 20(12):1311-1317.
DOI: 10.11805/TKYDA2021392
Abstract:
A communication signal modulation recognition method based on Transformer model is proposed. In the data preparation stage, a Different Symbol Rate Modulation Recognition(DSRMR) data set is constructed. In the data preprocessing stage, a method of I/Q data enhancement is proposed to meet the quantitative and diverse requirements of model training, and to enhance the generalization ability of the model. In the model construction stage, the method of slice serialization is introduced into the modulation recognition Transformer model, and it is employed to optimize the input problem of the Transformer neural network model. Experimental results prove that the communication signal modulation recognition method based on the Transformer model can obtain high-precision in signal automatic modulation recognition.
A communication signal modulation recognition method based on Transformer model is proposed. In the data preparation stage, a Different Symbol Rate Modulation Recognition(DSRMR) data set is constructed. In the data preprocessing stage, a method of I/Q data enhancement is proposed to meet the quantitative and diverse requirements of model training, and to enhance the generalization ability of the model. In the model construction stage, the method of slice serialization is introduced into the modulation recognition Transformer model, and it is employed to optimize the input problem of the Transformer neural network model. Experimental results prove that the communication signal modulation recognition method based on the Transformer model can obtain high-precision in signal automatic modulation recognition.
2022, 20(12):1326-1334.
DOI: 10.11805/TKYDA2022036
Abstract:
Signal modulation identification technology has important applications in both civilian and military fields. In the current information battlefield, due to the increasing number of information radiation sources such as various radars, communications, navigation, and electronic warfare weapons, the modulation forms are becoming more and more diverse, and the signal density is increasing, which makes the electromagnetic environment of war increasingly complicated, therefore the traditional signal modulation identification technology has been unable to adapt. A robust feature extraction, fusion and recognition technology of complex communication modulation signals is put forward, and a deep learning-based AlexNet network and complex neural network are proposed. Multimodal information in the statistical graph domain and signal I/Q waveform domain is fused for signal modulation identification. The simulation results show that the recognition accuracy of the proposed method is higher than that of the single-modal recognition method and the method without the multi-modal collaborative fusion framework under different Signal-to-Noise Ratios(SNRs).
Signal modulation identification technology has important applications in both civilian and military fields. In the current information battlefield, due to the increasing number of information radiation sources such as various radars, communications, navigation, and electronic warfare weapons, the modulation forms are becoming more and more diverse, and the signal density is increasing, which makes the electromagnetic environment of war increasingly complicated, therefore the traditional signal modulation identification technology has been unable to adapt. A robust feature extraction, fusion and recognition technology of complex communication modulation signals is put forward, and a deep learning-based AlexNet network and complex neural network are proposed. Multimodal information in the statistical graph domain and signal I/Q waveform domain is fused for signal modulation identification. The simulation results show that the recognition accuracy of the proposed method is higher than that of the single-modal recognition method and the method without the multi-modal collaborative fusion framework under different Signal-to-Noise Ratios(SNRs).
2022, 20(12):1335-1342.
DOI: 10.11805/TKYDA2022044
Abstract:
Spectrum data contains a large number of signals in radio monitoring. Accurate extraction of these signals is conducive to mastering the spectrum usage of the whole frequency band. Due to the interference of noise, several energy values of frequency points in the signal spectrum bandwidth will be lower than the detection threshold, then the traditional threshold detection algorithm will misestimate the signal as multiple signals and generate multiple false adjacent signal intervals, resulting in a decline of the spectrum signal extraction accuracy. To tackle this problem, an algorithm of spectrum signal extraction for estimating the signal number adaptively is proposed according to the characteristics of false adjacent signal intervals. The new method can estimate the number of electromagnetic signals in the spectrum monitoring data, and extract the corresponding signal and spectrum information accurately and automatically. The experimental result shows that the new method is adaptive, strongly robust and accurate, effectively improves the accuracy of spectrum signal extraction. It can provide basic electromagnetic signal data for supporting the identification of electromagnetic environment in military and civil spectrum monitoring.
Spectrum data contains a large number of signals in radio monitoring. Accurate extraction of these signals is conducive to mastering the spectrum usage of the whole frequency band. Due to the interference of noise, several energy values of frequency points in the signal spectrum bandwidth will be lower than the detection threshold, then the traditional threshold detection algorithm will misestimate the signal as multiple signals and generate multiple false adjacent signal intervals, resulting in a decline of the spectrum signal extraction accuracy. To tackle this problem, an algorithm of spectrum signal extraction for estimating the signal number adaptively is proposed according to the characteristics of false adjacent signal intervals. The new method can estimate the number of electromagnetic signals in the spectrum monitoring data, and extract the corresponding signal and spectrum information accurately and automatically. The experimental result shows that the new method is adaptive, strongly robust and accurate, effectively improves the accuracy of spectrum signal extraction. It can provide basic electromagnetic signal data for supporting the identification of electromagnetic environment in military and civil spectrum monitoring.
2022, 20(12):1343-1349.
DOI: 10.11805/TKYDA2022030
Abstract:
In the context of Cognitive Radio(CR), dynamic spectrum access has become a key approach to improve the spectrum utilization in wireless network. In this paper, using fine-grained spectrum measurement data collected from Global System for Mobile Communications-Railway(GSM-R), a data-driven deep-learning method is proposed to model the spectrum pattern and a framework is developed for dynamic spectrum access. A deep spectrum generative model is adopted to guide the spectrum allocation. A deep network that combines recurrent series model and background feature embeddings is developed to model and predict the short-term spectrum occupancy, then a strategy is proposed for dynamic channel access. Furthermore, a frequency-hopping system is implemented with Software Defined Radio(SDR) platform and it is integrated with the proposed strategy. The throughput capacity of this system is evaluated with real-world historical spectrum data. It is shown that the proposed method and system can enhance the ability of opportunistic communication and utilize the spectrum resource efficiently. The proposed spectrum access framework and the implementation with SDR are of great generality, so that they can be easily integrated into systems with different scenarios and frequency spans.
In the context of Cognitive Radio(CR), dynamic spectrum access has become a key approach to improve the spectrum utilization in wireless network. In this paper, using fine-grained spectrum measurement data collected from Global System for Mobile Communications-Railway(GSM-R), a data-driven deep-learning method is proposed to model the spectrum pattern and a framework is developed for dynamic spectrum access. A deep spectrum generative model is adopted to guide the spectrum allocation. A deep network that combines recurrent series model and background feature embeddings is developed to model and predict the short-term spectrum occupancy, then a strategy is proposed for dynamic channel access. Furthermore, a frequency-hopping system is implemented with Software Defined Radio(SDR) platform and it is integrated with the proposed strategy. The throughput capacity of this system is evaluated with real-world historical spectrum data. It is shown that the proposed method and system can enhance the ability of opportunistic communication and utilize the spectrum resource efficiently. The proposed spectrum access framework and the implementation with SDR are of great generality, so that they can be easily integrated into systems with different scenarios and frequency spans.
2022, 20(12):1269-1276.
DOI: 10.11805/TKYDA2021148
Abstract:
The increasingly crowded and complex airspace environment makes it necessary to determine the initiation of the true target track. Most existing research on radar target track initiation only considers one of real-time or initiation rate, and it is difficult to complete fast and accurate track initiation in a strong clutter environment. In this paper, a track initiation algorithm is proposed based on Deep Learning and Temporal-Spatial(DLTS) characteristics of radar measurement suitable for strong clutter environment. The algorithm first selects the candidate set from the radar measurement combinations, and next extracts the temporal change vector and spatial distribution vector of the measurement combination, and uses them as the input of the One-dimensional Convolutional Neural Network(1DCNN) and Gated Recurrent Unit(GRU) hybrid model to obtain the time dimensional characteristics and space dimensional characteristics of the measurement combination, then merge the two to get the temporal-spatial characteristics. Finally, the true and false tracks are classified with the temporal-spatial characteristics processed by self-attention, and the track initiation is completed. The simulations show that DLTS algorithm can effectively improve the performance of the true and false track initiation rate when the time loss is similar to that of the logic method in the strong clutter environment.
The increasingly crowded and complex airspace environment makes it necessary to determine the initiation of the true target track. Most existing research on radar target track initiation only considers one of real-time or initiation rate, and it is difficult to complete fast and accurate track initiation in a strong clutter environment. In this paper, a track initiation algorithm is proposed based on Deep Learning and Temporal-Spatial(DLTS) characteristics of radar measurement suitable for strong clutter environment. The algorithm first selects the candidate set from the radar measurement combinations, and next extracts the temporal change vector and spatial distribution vector of the measurement combination, and uses them as the input of the One-dimensional Convolutional Neural Network(1DCNN) and Gated Recurrent Unit(GRU) hybrid model to obtain the time dimensional characteristics and space dimensional characteristics of the measurement combination, then merge the two to get the temporal-spatial characteristics. Finally, the true and false tracks are classified with the temporal-spatial characteristics processed by self-attention, and the track initiation is completed. The simulations show that DLTS algorithm can effectively improve the performance of the true and false track initiation rate when the time loss is similar to that of the logic method in the strong clutter environment.
2022, 20(12):1249-1256.
DOI: 10.11805/TKYDA2021271
Abstract:
For few-shot electromagnetic signal classification, data augmentation is the most intuitive strategy. In this paper, Generative Adversarial Network(GAN) is employed to generate fake signal samples. The coarse-grained and fine-grained screening mechanisms are designed to screen the generated fake signals. The generated signals with poor quality are removed and the effective expansion of training dataset is realized. In order to verify the effectiveness of the proposed data augmentation algorithm, sufficient experiments are conducted on the RADIOML 2016.04C dataset. Experimental results show that the proposed method can improve the accuracy of few-shot electromagnetic signal classification effectively.
For few-shot electromagnetic signal classification, data augmentation is the most intuitive strategy. In this paper, Generative Adversarial Network(GAN) is employed to generate fake signal samples. The coarse-grained and fine-grained screening mechanisms are designed to screen the generated fake signals. The generated signals with poor quality are removed and the effective expansion of training dataset is realized. In order to verify the effectiveness of the proposed data augmentation algorithm, sufficient experiments are conducted on the RADIOML 2016.04C dataset. Experimental results show that the proposed method can improve the accuracy of few-shot electromagnetic signal classification effectively.
2022, 20(12):1277-1284.
DOI: 10.11805/TKYDA2021178
Abstract:
In recent years, the research on risk measurement and control methods based on big data analysis model has become more and more important, and the backtesting analysis on risk measurement tools can guarantee the effectiveness of the techniques used in actual data analysis. Marginal Expected Shortfall(MES) is an important tool to measure the marginal contribution of individual institutions to systemic risk, and the backtest methodologies for MES is also worthy to focus on. In this paper, the backtest method of ES in C. Acerb et al. is extended to the two-dimensional case and two backtest methodologies are proposed for MES. The results of simulation show that these two new statistics are more powerful than the statistics used in D.Banulescu et al. under situations that the difference between the null hypothesis and the alternative hypothesis is relatively small. The results of empirical analysis also support that these two new statistics proposed in this paper accept the null hypothesis more cautiously under the same prediction model. This method can give some reference for model algorithm backtesing under big data.
In recent years, the research on risk measurement and control methods based on big data analysis model has become more and more important, and the backtesting analysis on risk measurement tools can guarantee the effectiveness of the techniques used in actual data analysis. Marginal Expected Shortfall(MES) is an important tool to measure the marginal contribution of individual institutions to systemic risk, and the backtest methodologies for MES is also worthy to focus on. In this paper, the backtest method of ES in C. Acerb et al. is extended to the two-dimensional case and two backtest methodologies are proposed for MES. The results of simulation show that these two new statistics are more powerful than the statistics used in D.Banulescu et al. under situations that the difference between the null hypothesis and the alternative hypothesis is relatively small. The results of empirical analysis also support that these two new statistics proposed in this paper accept the null hypothesis more cautiously under the same prediction model. This method can give some reference for model algorithm backtesing under big data.
2022, 20(12):1292-1297.
DOI: 10.11805/TKYDA2021270
Abstract:
The increasingly complex electromagnetic environment imposes high requirements for battlefield target detection. Accurate, quick and complete multi-radar track correlation has become an urgent problem with the continuous development of multi-radar fusion systems. Most of the existing research on track correlation only considers the latest target track points reported by radar, while ignoring the previous track information. In addition, the solution to the asynchronous track problem of most track correlation algorithms is time registration. It not only increases the computational cost of the algorithm, but also magnifies the error contained in the track information. Therefore, time registration is difficult to be applied to the current complex electromagnetic environment. In this paper, a Track-to-Track Correlation algorithm based on Siamese Network(TTCSN) is proposed, which is suitable for asynchronous track correlation and does not need time registration. A pair of tracks are sent into the feature extraction network, and TTCSN learns the hidden features of input tracks. Then the similarity of hidden feature vectors are calculated by TTCSN to get the similarity vector which is fed into the classifier to distinguish that the input tracks are correlated or not. The experimental results show that TTCSN algorithm can effectively solve the problem of asynchronous track correlation.
The increasingly complex electromagnetic environment imposes high requirements for battlefield target detection. Accurate, quick and complete multi-radar track correlation has become an urgent problem with the continuous development of multi-radar fusion systems. Most of the existing research on track correlation only considers the latest target track points reported by radar, while ignoring the previous track information. In addition, the solution to the asynchronous track problem of most track correlation algorithms is time registration. It not only increases the computational cost of the algorithm, but also magnifies the error contained in the track information. Therefore, time registration is difficult to be applied to the current complex electromagnetic environment. In this paper, a Track-to-Track Correlation algorithm based on Siamese Network(TTCSN) is proposed, which is suitable for asynchronous track correlation and does not need time registration. A pair of tracks are sent into the feature extraction network, and TTCSN learns the hidden features of input tracks. Then the similarity of hidden feature vectors are calculated by TTCSN to get the similarity vector which is fed into the classifier to distinguish that the input tracks are correlated or not. The experimental results show that TTCSN algorithm can effectively solve the problem of asynchronous track correlation.
