Email alert
2022 Vol. 44, No. 6
Display Method:
2022, 44(6): 1879-1886.
doi: 10.11999/JEIT211343
Abstract:
In underwater acoustic channel equalization, the channel estimation-based equalization has better performance theoretically, but the high computational complexity limits its practical applications. To solve this problem, an iterative Kalman equalizer based on Kalman filter and Turbo equalization is proposed firstly, which realizes iterative channel estimation and iterative Kalman equalization based on soft symbols generated by the channel decoder, and the complexity is about one order of magnitude lower than that of conventional methods. Secondly, aiming at the error transmission of a single equalization algorithm and single direction Turbo equalizer structure, a hybrid bi-directional Turbo equalizer based on iterative Kalman equalizer and Improved Proportional Normalized LMS (IPNLMS) adaptive equalizer is designed, which improves the convergence speed and equalization performance of the adaptive equalizer, and reduces the error transmission through bi-directional equalization gain. The proposed hybrid bi-directional Turbo equalization for underwater acoustic communications based on the Kalman filter is verified by theoretical analysis and simulation.
In underwater acoustic channel equalization, the channel estimation-based equalization has better performance theoretically, but the high computational complexity limits its practical applications. To solve this problem, an iterative Kalman equalizer based on Kalman filter and Turbo equalization is proposed firstly, which realizes iterative channel estimation and iterative Kalman equalization based on soft symbols generated by the channel decoder, and the complexity is about one order of magnitude lower than that of conventional methods. Secondly, aiming at the error transmission of a single equalization algorithm and single direction Turbo equalizer structure, a hybrid bi-directional Turbo equalizer based on iterative Kalman equalizer and Improved Proportional Normalized LMS (IPNLMS) adaptive equalizer is designed, which improves the convergence speed and equalization performance of the adaptive equalizer, and reduces the error transmission through bi-directional equalization gain. The proposed hybrid bi-directional Turbo equalization for underwater acoustic communications based on the Kalman filter is verified by theoretical analysis and simulation.
2022, 44(6): 1887-1896.
doi: 10.11999/JEIT211284
Abstract:
Conventional Direction Of Arrival (DOA) estimators achieve satisfactory performance with the common assumptions of Gaussian noise. However, the impulsive noise exists in the shallow water extensively and does not follow the Gaussian distribution, which induce undesirable biases and degrade the performance of the conventional estimators. In the paper, a new DOA estimation method based on variational Bayesian inference in presence of shallow water non-Gaussian noise is proposed to improve the DOA estimation performance. Firstly, the multiple measurement vectors Sparse Signal Representation (SSR) model is formulated utilizing the sparsity of signal and impulsive noise. After that, the hierarchical Bayesian estimation framework is formulated which considers the common sparsity of signal and the independent sparsity of impulsive noise. Subsequently, the variational Bayesian inference is utilized to achieve the posterior estimations for the signal and impulsive noise. The SSR model incorporates the off-grid bias, and the root sparse Bayesian learning realizes to refine the bias and mitigate the basis mismatches. At last, the accurate DOA estimation is achieved through iterative updates and the effects of impulsive noise are mitigated. Simulations are used to verify that the proposed estimator achieves superior performance compared with state-of-art benchmarks.
Conventional Direction Of Arrival (DOA) estimators achieve satisfactory performance with the common assumptions of Gaussian noise. However, the impulsive noise exists in the shallow water extensively and does not follow the Gaussian distribution, which induce undesirable biases and degrade the performance of the conventional estimators. In the paper, a new DOA estimation method based on variational Bayesian inference in presence of shallow water non-Gaussian noise is proposed to improve the DOA estimation performance. Firstly, the multiple measurement vectors Sparse Signal Representation (SSR) model is formulated utilizing the sparsity of signal and impulsive noise. After that, the hierarchical Bayesian estimation framework is formulated which considers the common sparsity of signal and the independent sparsity of impulsive noise. Subsequently, the variational Bayesian inference is utilized to achieve the posterior estimations for the signal and impulsive noise. The SSR model incorporates the off-grid bias, and the root sparse Bayesian learning realizes to refine the bias and mitigate the basis mismatches. At last, the accurate DOA estimation is achieved through iterative updates and the effects of impulsive noise are mitigated. Simulations are used to verify that the proposed estimator achieves superior performance compared with state-of-art benchmarks.
2022, 44(6): 1897-1905.
doi: 10.11999/JEIT211333
Abstract:
The mismatch of sound speed induced by internal solitary waves will cause the inaccurate estimation in the matched field processing for source localization. In this paper, a robust Adaptive Matched Field Processing method of Rank Reduction (RR-AMFP) for internal solitary waves is proposed. Based on the traditional adaptive matched field processing algorithm, this method integrates dominant mode rejection beamforming, and reduces the rank of sampling covariance matrix by eigen-decomposition, and suppresses the noise space. Meanwhile, a suppressing coefficient and a weighting factor are used to calculate the weight vector in the matching process, and the mismatched coping vectors are detected. Therefore, this method can maintain better robustness in the internal solitary wave environment, and the reduction of rank also shortens the calculation time. The simulation results show that this method can accurately estimate the source location under a single internal solitary wave, but the internal solitary wave train with large amplitude will still lead to more errors of estimation. The estimated distance error is 3.3% and depth error is 1.5% in the localization experiment of internal solitary waves in the South China Sea, which belongs to reliable localization. The experiment results demonstrate the effectiveness of the method in the actual environment with internal solitary waves.
The mismatch of sound speed induced by internal solitary waves will cause the inaccurate estimation in the matched field processing for source localization. In this paper, a robust Adaptive Matched Field Processing method of Rank Reduction (RR-AMFP) for internal solitary waves is proposed. Based on the traditional adaptive matched field processing algorithm, this method integrates dominant mode rejection beamforming, and reduces the rank of sampling covariance matrix by eigen-decomposition, and suppresses the noise space. Meanwhile, a suppressing coefficient and a weighting factor are used to calculate the weight vector in the matching process, and the mismatched coping vectors are detected. Therefore, this method can maintain better robustness in the internal solitary wave environment, and the reduction of rank also shortens the calculation time. The simulation results show that this method can accurately estimate the source location under a single internal solitary wave, but the internal solitary wave train with large amplitude will still lead to more errors of estimation. The estimated distance error is 3.3% and depth error is 1.5% in the localization experiment of internal solitary waves in the South China Sea, which belongs to reliable localization. The experiment results demonstrate the effectiveness of the method in the actual environment with internal solitary waves.
2022, 44(6): 1906-1918.
doi: 10.11999/JEIT211371
Abstract:
Focusing on solving the depth discrimination of negative thermocline, an approach based on the slope distribution of interference pattern is present. In this method, a model for the variation of interference pattern distribution of radiated noise with the source depth is established using the difference in the excitation ability of surface and underwater sources of normal mode in shallow water waveguides with negative thermocline, and the physical mechanism of the difference in the slope distribution of interference pattern of radiated noise excited by surface and underwater sources is analyzed. Using the image processing algorithm, the difference is presented as the number of peaks in the column variance vector of the Radon transform matrix of the interference image of radiated noise, and the surface and underwater sources are discriminated accordingly. Simulative and experimental results show that the method can effectively discriminate between surface and underwater sources in shallow water waveguide with negative thermocline, and doesn’t require a priori information about the source distance and ocean acoustic environment parameters compared with the conventional method.
Focusing on solving the depth discrimination of negative thermocline, an approach based on the slope distribution of interference pattern is present. In this method, a model for the variation of interference pattern distribution of radiated noise with the source depth is established using the difference in the excitation ability of surface and underwater sources of normal mode in shallow water waveguides with negative thermocline, and the physical mechanism of the difference in the slope distribution of interference pattern of radiated noise excited by surface and underwater sources is analyzed. Using the image processing algorithm, the difference is presented as the number of peaks in the column variance vector of the Radon transform matrix of the interference image of radiated noise, and the surface and underwater sources are discriminated accordingly. Simulative and experimental results show that the method can effectively discriminate between surface and underwater sources in shallow water waveguide with negative thermocline, and doesn’t require a priori information about the source distance and ocean acoustic environment parameters compared with the conventional method.
2022, 44(6): 1919-1926.
doi: 10.11999/JEIT211359
Abstract:
The cooperative formation of Autonomous Underwater Vehicles (AUVs) can be used to locate targets in unknown waters. In order to reduce the positioning error caused by AUV navigation errors, a multi-AUV formation fusion observation cooperative target positioning algorithm based on Extended Kalman Filter (EKF) is proposed in this paper. The AUV formation consists of a pilot AUV equipped with a high precision Inertial Navigation System (INS) and several following AUVs equipped with low precision INS. The reference AUV and the AUV to be measured are selected from the following AUV, and different observations are made respectively by setting the positioning period and observation interval. After receiving position parameters from high-precision AUV with reference to AUV as relay, its own position parameters are transmitted to the AUV to be tested, and cooperative position correction of AUV cluster is completed with extended Kalman filter. Simulation results show that the AUV cluster has high positioning accuracy and small error accumulation over time, which requires less piloting AUV quantity and can achieve low power consumption and long-distance positioning of underwater targets.
The cooperative formation of Autonomous Underwater Vehicles (AUVs) can be used to locate targets in unknown waters. In order to reduce the positioning error caused by AUV navigation errors, a multi-AUV formation fusion observation cooperative target positioning algorithm based on Extended Kalman Filter (EKF) is proposed in this paper. The AUV formation consists of a pilot AUV equipped with a high precision Inertial Navigation System (INS) and several following AUVs equipped with low precision INS. The reference AUV and the AUV to be measured are selected from the following AUV, and different observations are made respectively by setting the positioning period and observation interval. After receiving position parameters from high-precision AUV with reference to AUV as relay, its own position parameters are transmitted to the AUV to be tested, and cooperative position correction of AUV cluster is completed with extended Kalman filter. Simulation results show that the AUV cluster has high positioning accuracy and small error accumulation over time, which requires less piloting AUV quantity and can achieve low power consumption and long-distance positioning of underwater targets.
2022, 44(6): 1927-1936.
doi: 10.11999/JEIT211432
Abstract:
Due to various effects, such as ocean currents, locations of sensor nodes have to be updated in Underwater Acoustic Sensors Networks (UASNs). In UASNs localization, using an Autonomous Underwater Vehicle (AUV) as the mobile anchor can reduce the localization cost. However, the energy utilization of AUV is not efficient. In order to improve the energy utilization of AUV, a dynamic path planning method is proposed for an AUV-aided localization for UASNs. In this method, the location correction process is regarded as a process of reducing the entropy of location information of sensor nodes. In dynamic path planning, the next target location of the AUV is planned according to the sensor node location information and the expected AUV energy consumption. The greedy algorithm is used to select the location that can obtain the maximum ratio of the expectation of the information gain and mobile energy consumption as the target location. The simulation show that the proposed algorithm can improve the energy efficiency while ensuring the positioning accuracy.
Due to various effects, such as ocean currents, locations of sensor nodes have to be updated in Underwater Acoustic Sensors Networks (UASNs). In UASNs localization, using an Autonomous Underwater Vehicle (AUV) as the mobile anchor can reduce the localization cost. However, the energy utilization of AUV is not efficient. In order to improve the energy utilization of AUV, a dynamic path planning method is proposed for an AUV-aided localization for UASNs. In this method, the location correction process is regarded as a process of reducing the entropy of location information of sensor nodes. In dynamic path planning, the next target location of the AUV is planned according to the sensor node location information and the expected AUV energy consumption. The greedy algorithm is used to select the location that can obtain the maximum ratio of the expectation of the information gain and mobile energy consumption as the target location. The simulation show that the proposed algorithm can improve the energy efficiency while ensuring the positioning accuracy.
2022, 44(6): 1937-1946.
doi: 10.11999/JEIT211447
Abstract:
When Mapping Sequences Spread Spectrum (MSSS) method is applied to the Parallel Combinatory Spread Spectrum (PCSS) UnderWater Acoustic (UWA) communication, the peak-to-mean envelope power ratio of the output signal is reduced, and hence the communication performance is improved. However, if Gold codes are selected as the spread sequence, due to the property of Gold codes, there will be a pseudo correlation peak when the mapping signal is correlated with the spread sequence set. Hence, the communication performance is degraded seriously. In order to overcome the performance degradation, two new schemes are proposed in this paper, namely the Parallel Combinatory Spread Spectrum based on Phase Differences of Correlation Peaks (PDCP-PCSS) and the Interleaved Parallel Combinatory Spread Spectrum (IPCSS) respectively. PDCP-PCSS can effectively identify the pseudo correlation peak and eliminate it. IPCSS is able to generate signal without the pseudo correlation peak by adding the interleavers in the transmitter. Simulation and sea trail results show that the performance of two proposed schemes in the paper outperform the Conventional Parallel Combinatory Spread Spectrum (CPCSS). Among them, the performance of PDCP-PCSS is the best. However it is only suitable for the combination of three Gold codes while IPCSS has less limitation in usage.
When Mapping Sequences Spread Spectrum (MSSS) method is applied to the Parallel Combinatory Spread Spectrum (PCSS) UnderWater Acoustic (UWA) communication, the peak-to-mean envelope power ratio of the output signal is reduced, and hence the communication performance is improved. However, if Gold codes are selected as the spread sequence, due to the property of Gold codes, there will be a pseudo correlation peak when the mapping signal is correlated with the spread sequence set. Hence, the communication performance is degraded seriously. In order to overcome the performance degradation, two new schemes are proposed in this paper, namely the Parallel Combinatory Spread Spectrum based on Phase Differences of Correlation Peaks (PDCP-PCSS) and the Interleaved Parallel Combinatory Spread Spectrum (IPCSS) respectively. PDCP-PCSS can effectively identify the pseudo correlation peak and eliminate it. IPCSS is able to generate signal without the pseudo correlation peak by adding the interleavers in the transmitter. Simulation and sea trail results show that the performance of two proposed schemes in the paper outperform the Conventional Parallel Combinatory Spread Spectrum (CPCSS). Among them, the performance of PDCP-PCSS is the best. However it is only suitable for the combination of three Gold codes while IPCSS has less limitation in usage.
2022, 44(6): 1947-1955.
doi: 10.11999/JEIT211407
Abstract:
The translation equivariance of convolutional layers are not compatible with the linear spectrum. Therefore, the convolutional networks can not carry the long-distance dependency of high-dimensional features. One bi-logarithmic spectrum feature is presented by this paper for classification of ship radiated noise. This bi-logarithmic spectrum rearranges the frequency points of the logarithmic spectrum to ensure the resolution of high frequencies, therefore the substantial deep convolutional network is not necessary. Considering on the prior knowledge that each row of the bi-logarithmic spectrum corresponding to the same one target, a convolutional network as well as an objective function are constructed. Then this network is trained and tested with DeepShip dataset to classify four types of marine vessels, and the results show that, with the same feature dimensions, the classification accuracy of the algorithm proposed by this paper is improved by 2.4% more than the convolutional network with the input feature of linear scale spectrum.
The translation equivariance of convolutional layers are not compatible with the linear spectrum. Therefore, the convolutional networks can not carry the long-distance dependency of high-dimensional features. One bi-logarithmic spectrum feature is presented by this paper for classification of ship radiated noise. This bi-logarithmic spectrum rearranges the frequency points of the logarithmic spectrum to ensure the resolution of high frequencies, therefore the substantial deep convolutional network is not necessary. Considering on the prior knowledge that each row of the bi-logarithmic spectrum corresponding to the same one target, a convolutional network as well as an objective function are constructed. Then this network is trained and tested with DeepShip dataset to classify four types of marine vessels, and the results show that, with the same feature dimensions, the classification accuracy of the algorithm proposed by this paper is improved by 2.4% more than the convolutional network with the input feature of linear scale spectrum.
2022, 44(6): 1956-1965.
doi: 10.11999/JEIT211374
Abstract:
In view of the weak ability of traditional Hidden Markov model (HMM) method to extract time-varying line spectrum and multi line spectrum and the large amount of calculation in dynamic programming process, a One- Dimensional Hidden Markov Model (1D-HMM) method based on dynamic parameters is proposed to extract line spectrum trajectory in LOw Frequency Analysis and Recording (LOFAR) diagram of underwater acoustic signal. In this method, the time-varying frequency state is modeled as a first-order Markov process, and the Viterbi algorithm is repeated several times to extract multiple frequency trajectories with the largest a posteriori probability. In the iterative process, the state transition probability matrix in HMM is dynamically adjusted by the first derivative of the sequence calculated in real time, which improves the extraction ability of line spectrum trajectory and the resolution ability of multi line spectrum; A power spectrum accumulation method based on dynamic sliding window is designed to estimate the birth and death of line spectrum and eliminate false line spectrum trajectories. At the same time, the block processing strategy is designed for LOFAR graph data in the implementation process, which reduces greatly the amount of calculation. The simulation and actual data processing results show that the method can effectively detect and track the frequency state of complex time-varying spectrum with good operational efficiency under low signal-to-noise ratio conditions, which provides good technical support for the detection of weak signals in sonar devices.
In view of the weak ability of traditional Hidden Markov model (HMM) method to extract time-varying line spectrum and multi line spectrum and the large amount of calculation in dynamic programming process, a One- Dimensional Hidden Markov Model (1D-HMM) method based on dynamic parameters is proposed to extract line spectrum trajectory in LOw Frequency Analysis and Recording (LOFAR) diagram of underwater acoustic signal. In this method, the time-varying frequency state is modeled as a first-order Markov process, and the Viterbi algorithm is repeated several times to extract multiple frequency trajectories with the largest a posteriori probability. In the iterative process, the state transition probability matrix in HMM is dynamically adjusted by the first derivative of the sequence calculated in real time, which improves the extraction ability of line spectrum trajectory and the resolution ability of multi line spectrum; A power spectrum accumulation method based on dynamic sliding window is designed to estimate the birth and death of line spectrum and eliminate false line spectrum trajectories. At the same time, the block processing strategy is designed for LOFAR graph data in the implementation process, which reduces greatly the amount of calculation. The simulation and actual data processing results show that the method can effectively detect and track the frequency state of complex time-varying spectrum with good operational efficiency under low signal-to-noise ratio conditions, which provides good technical support for the detection of weak signals in sonar devices.
2022, 44(6): 1966-1973.
doi: 10.11999/JEIT220048
Abstract:
The goal of this paper is to design a routing protocol for IoUT data delivery. It is well known that underwater communication is limited by its channel’s hazardous nature, thus efficient routing protocols are needed to compensate for the challenging environments. Geometric STAteless Routing (G-STAR) is a type of geographical routing protocol that forwards messages in a greedy way and obtains sub-optimal results in most 3-dimensional Internet of Things scenarios. Yet the numerous degrading factors in underwater channels limit severely the performance of the G-STAR protocol. In this paper, a Hybrid G-STAR (H-G-STAR) routing protocol continuing the advantage of G-STAR and specifically adapted for underwater conditions is proposed. By introducing a noncooperative routing tactic based on channel listening, the protocol spontaneously searches for the route with the best channel condition whenever the network condition permits and thus avoiding inferior channels greedy routing might encounter. The simulation results show that the proposed protocol improves the routing performance of the network, obtains a lower Bit Error Rate (BER) in the physical layer than G-STAR and is better adapted for underwater network topologies.
The goal of this paper is to design a routing protocol for IoUT data delivery. It is well known that underwater communication is limited by its channel’s hazardous nature, thus efficient routing protocols are needed to compensate for the challenging environments. Geometric STAteless Routing (G-STAR) is a type of geographical routing protocol that forwards messages in a greedy way and obtains sub-optimal results in most 3-dimensional Internet of Things scenarios. Yet the numerous degrading factors in underwater channels limit severely the performance of the G-STAR protocol. In this paper, a Hybrid G-STAR (H-G-STAR) routing protocol continuing the advantage of G-STAR and specifically adapted for underwater conditions is proposed. By introducing a noncooperative routing tactic based on channel listening, the protocol spontaneously searches for the route with the best channel condition whenever the network condition permits and thus avoiding inferior channels greedy routing might encounter. The simulation results show that the proposed protocol improves the routing performance of the network, obtains a lower Bit Error Rate (BER) in the physical layer than G-STAR and is better adapted for underwater network topologies.
2022, 44(6): 1974-1983.
doi: 10.11999/JEIT211077
Abstract:
In recent years, ship target recognition based on machine learning has become an important research direction in the field of underwater acoustic signal processing, but the acquisition of underwater acoustic target signal is difficult, and the problem of insufficient sample size and imbalance leads easily to the poor recognition effect of target classification model. A ship noise data classification method based on Generative Admission-Network (GAN) is proposed in this paper. This method uses generative admission-learning theory to generate pseudo-DEMON modulation spectrum data with stronger nonlinear characteristics and richer feature differences compared with traditional data enhancement algorithms to alleviate the problem of insufficient training sample size. Then, the output of the whole connection layer in the traditional generative adversarial network is replaced by an ensemble classifier which is better at solving the problem of small samples, so as to reduce the dependence of the classifier on the amount of data and improve further the performance of the classification model. Finally, experimental results based on real samples show that, compared with traditional data enhancement algorithms and generative adversarial networks, the proposed method can improve the classification performance of models with insufficient samples more effectively.
In recent years, ship target recognition based on machine learning has become an important research direction in the field of underwater acoustic signal processing, but the acquisition of underwater acoustic target signal is difficult, and the problem of insufficient sample size and imbalance leads easily to the poor recognition effect of target classification model. A ship noise data classification method based on Generative Admission-Network (GAN) is proposed in this paper. This method uses generative admission-learning theory to generate pseudo-DEMON modulation spectrum data with stronger nonlinear characteristics and richer feature differences compared with traditional data enhancement algorithms to alleviate the problem of insufficient training sample size. Then, the output of the whole connection layer in the traditional generative adversarial network is replaced by an ensemble classifier which is better at solving the problem of small samples, so as to reduce the dependence of the classifier on the amount of data and improve further the performance of the classification model. Finally, experimental results based on real samples show that, compared with traditional data enhancement algorithms and generative adversarial networks, the proposed method can improve the classification performance of models with insufficient samples more effectively.
2022, 44(6): 1984-1990.
doi: 10.11999/JEIT210949
Abstract:
When the Index Modulated Filter Bank MultiCarrier (FBMC-IM) underwater acoustic communication system carries out signal detection, the first step is to determine the index of the active subcarriers according to the recovered data after equalization. In this paper, the advantage of Bidirectional Long Short-Term Memory (BLSTM) network for feature extraction of chronological signals is combined, the deep learning theory is introduced into the concept of underwater acoustic signal processing, and an index detection method based on deep BLSTM is proposed. The improved algorithm can increase the estimation accuracy by transforming the index detection into a data-driven multivariate classification. Compared with the traditional methods, the proposed algorithm has lower computational complexity but better bit error ratio performance. The superiority and robustness of the proposed method are verified by the simulation based on lake trial channel data, which can be considered as a general detection method under index modulation mechanism.
When the Index Modulated Filter Bank MultiCarrier (FBMC-IM) underwater acoustic communication system carries out signal detection, the first step is to determine the index of the active subcarriers according to the recovered data after equalization. In this paper, the advantage of Bidirectional Long Short-Term Memory (BLSTM) network for feature extraction of chronological signals is combined, the deep learning theory is introduced into the concept of underwater acoustic signal processing, and an index detection method based on deep BLSTM is proposed. The improved algorithm can increase the estimation accuracy by transforming the index detection into a data-driven multivariate classification. Compared with the traditional methods, the proposed algorithm has lower computational complexity but better bit error ratio performance. The superiority and robustness of the proposed method are verified by the simulation based on lake trial channel data, which can be considered as a general detection method under index modulation mechanism.
2022, 44(6): 1991-1998.
doi: 10.11999/JEIT211445
Abstract:
Arctic underwater channels are complex. Matched field processing is a kind of target localizing method that can combine environmental information with signal processing methods. The target position can be accurately estimated when model is matched. However, due to the complex environment and the lack of key information, it is difficult to match the model completely, which leads to the problem of inaccurate estimation. In this paper, an improved matched field algorithm is proposed. Based on the principle of phase normalization, the sparse model is optimized, and the normalized information of arrays is used for matching to eliminate the influence of signal spectrum. The simulation result shows that the accuracy of the improved method is higher than that of the former method. The experimental data processing shows that the improved method has less error in target position estimation and higher resolution than the conventional matched field processing method.
Arctic underwater channels are complex. Matched field processing is a kind of target localizing method that can combine environmental information with signal processing methods. The target position can be accurately estimated when model is matched. However, due to the complex environment and the lack of key information, it is difficult to match the model completely, which leads to the problem of inaccurate estimation. In this paper, an improved matched field algorithm is proposed. Based on the principle of phase normalization, the sparse model is optimized, and the normalized information of arrays is used for matching to eliminate the influence of signal spectrum. The simulation result shows that the accuracy of the improved method is higher than that of the former method. The experimental data processing shows that the improved method has less error in target position estimation and higher resolution than the conventional matched field processing method.
2022, 44(6): 1999-2005.
doi: 10.11999/JEIT211128
Abstract:
To solve the lack of model switching and tracking accuracy of the existing Adaptive Interacting Multiple Model (AIMM) in the underwater target tracking, combined with the Unscented Kalman Filter, an improved AIMM-UKF algorithm is proposed. On the basis of adaptively modifying the Markov probability transition matrix, this algorithm uses the decision window to modify it twice to increase the probability of the matching model observably and reduce the effects of the mismatch model. Simulation results show that compared with the original adaptive algorithm, the improved algorithm can make fuller use of posterior information, has a better model switching speed, and improves tracking accuracy by about 24%.
To solve the lack of model switching and tracking accuracy of the existing Adaptive Interacting Multiple Model (AIMM) in the underwater target tracking, combined with the Unscented Kalman Filter, an improved AIMM-UKF algorithm is proposed. On the basis of adaptively modifying the Markov probability transition matrix, this algorithm uses the decision window to modify it twice to increase the probability of the matching model observably and reduce the effects of the mismatch model. Simulation results show that compared with the original adaptive algorithm, the improved algorithm can make fuller use of posterior information, has a better model switching speed, and improves tracking accuracy by about 24%.
2022, 44(6): 2006-2013.
doi: 10.11999/JEIT211398
Abstract:
Due to the serious interference in the ocean, underwater acoustic communications are very difficult. Spread spectrum communication technology has good anti-interference performance, and it can ensure reliable communications in complex marine environment, it is often used in underwater acoustic communications. Underwater acoustic channels are typical coherent multipath channels. The signals arriving along different paths have different propagation delays and angles of arrival, so received signals have space-time clustering characteristic. That is, received signals have time delay spread and angle spread. The coherent superposition of multipath signals leads to serious inter symbol interference in received signals. In order to take advantage of the space-time clustering characteristic of the underwater acoustic signals, a space-time processor is designed to filter the signals arriving along each path respectively. The reliability of the communication system can be effectively improved by combining the diversity of space-time clusters. An underwater acoustic spread spectrum communication scheme is proposed based on space-time cluster processing. This communication scheme is compared and analyzed in the simulation and the experiment, to verify its performance advantages.
Due to the serious interference in the ocean, underwater acoustic communications are very difficult. Spread spectrum communication technology has good anti-interference performance, and it can ensure reliable communications in complex marine environment, it is often used in underwater acoustic communications. Underwater acoustic channels are typical coherent multipath channels. The signals arriving along different paths have different propagation delays and angles of arrival, so received signals have space-time clustering characteristic. That is, received signals have time delay spread and angle spread. The coherent superposition of multipath signals leads to serious inter symbol interference in received signals. In order to take advantage of the space-time clustering characteristic of the underwater acoustic signals, a space-time processor is designed to filter the signals arriving along each path respectively. The reliability of the communication system can be effectively improved by combining the diversity of space-time clusters. An underwater acoustic spread spectrum communication scheme is proposed based on space-time cluster processing. This communication scheme is compared and analyzed in the simulation and the experiment, to verify its performance advantages.
2022, 44(6): 2014-2023.
doi: 10.11999/JEIT211274
Abstract:
Underwater wireless communication mainly relies on underwater acoustic communication for information transmission. The high propagation delay and high bit error rate of underwater acoustic link, however, is a challenging task to provide low delay communication services for underwater applications. The Coordinate Radio-Acoustic Network (CRAN) aims to utilize fully the on-surface radio link to mitigate the shortcoming of Underwater Acoustic Network (UAN) and improve the performances of whole network. The routing protocol of CRAN needs to construct the heterogeneous acoustic-radio links which is one of the key researches to CRAN. In this paper, first, the design of buoy node and internet-stack with its’ implementation in NS-3 are introduced, and the simulation platform of CRAN in Network Simulator 3 (NS-3) is built. Second, the application of reactive routing protocol as Ad-hoc On-demand Distance Vector routing (AODV) to CRAN is discussed, and it is found that AODV used distance vector could select more high-speed radio links during route construction, which could effectively reduce the network transmission delay. Finally, the performance of AODV and its counterpart are compared and analyzed through the simulation. The results show that, compared with underwater acoustic communication network, CRAN has a great improvement in transmission delay, packet delivery rate, network throughput, energy efficiency and routing response speed. At the same time, the reactive routing protocol represented by AODV is more suitable than the active routing protocol represented by Optimized Link State Routing (OLSR) for CRAN.
Underwater wireless communication mainly relies on underwater acoustic communication for information transmission. The high propagation delay and high bit error rate of underwater acoustic link, however, is a challenging task to provide low delay communication services for underwater applications. The Coordinate Radio-Acoustic Network (CRAN) aims to utilize fully the on-surface radio link to mitigate the shortcoming of Underwater Acoustic Network (UAN) and improve the performances of whole network. The routing protocol of CRAN needs to construct the heterogeneous acoustic-radio links which is one of the key researches to CRAN. In this paper, first, the design of buoy node and internet-stack with its’ implementation in NS-3 are introduced, and the simulation platform of CRAN in Network Simulator 3 (NS-3) is built. Second, the application of reactive routing protocol as Ad-hoc On-demand Distance Vector routing (AODV) to CRAN is discussed, and it is found that AODV used distance vector could select more high-speed radio links during route construction, which could effectively reduce the network transmission delay. Finally, the performance of AODV and its counterpart are compared and analyzed through the simulation. The results show that, compared with underwater acoustic communication network, CRAN has a great improvement in transmission delay, packet delivery rate, network throughput, energy efficiency and routing response speed. At the same time, the reactive routing protocol represented by AODV is more suitable than the active routing protocol represented by Optimized Link State Routing (OLSR) for CRAN.
2022, 44(6): 2024-2034.
doi: 10.11999/JEIT211311
Abstract:
To deal with the problem of frequency offset and carrier phase fluctuation caused by the conventional Doppler on the DSSS (Direct-Sequence Spread-Spectrum) underwater acoustic communication between AUVs (Autonomous Underwater Vehicles), a combination differential direct sequence spread spectrum algorithm is proposed. The Doppler rough estimation is realized through Two-dimension Combination Difference(TCD) spread spectrum frame structure, and the Doppler tolerance is extended through the Frequency Compression-Energy Detector(FC-ED), then an algorithm of combination differential is proposed for polarity decoding. Using the combination differential direct sequence spread spectrum algorithm, good performance (BEK<10–2) is achieved for a signal-to-noise ratio as low as –8 dB based on at-sea data.
To deal with the problem of frequency offset and carrier phase fluctuation caused by the conventional Doppler on the DSSS (Direct-Sequence Spread-Spectrum) underwater acoustic communication between AUVs (Autonomous Underwater Vehicles), a combination differential direct sequence spread spectrum algorithm is proposed. The Doppler rough estimation is realized through Two-dimension Combination Difference(TCD) spread spectrum frame structure, and the Doppler tolerance is extended through the Frequency Compression-Energy Detector(FC-ED), then an algorithm of combination differential is proposed for polarity decoding. Using the combination differential direct sequence spread spectrum algorithm, good performance (BEK<10–2) is achieved for a signal-to-noise ratio as low as –8 dB based on at-sea data.
2022, 44(6): 2035-2044.
doi: 10.11999/JEIT210348
Abstract:
To overcome disadvantages of Orthogonal Frequency Division Multiplexing(OFDM) underwater acoustic mobile communication, a data variance-based Doppler frequency shift estimation method is proposed to estimate the time-varying Doppler shifts. The estimated channel response of previous OFDM symbol is utilized to recover the valid data sequence and its frequency-domain diversity. The variance of the ratio of data sequence and diversity copy are calculated under different Doppler compensation factors. Then the correct Doppler frequency shift factor is achieved by seeking the minimum of the variance. Sparse Bayesian learning and decision feedback channel estimation algorithm are used for calculating the revised channel frequency-domain response. The channel response is propagated to the follow-up symbol to track the time-varying Doppler shifts. The feasibility and superiority of the proposed method are verified by simulation. The sea trail demonstrates that the proposed algorithm can effectively estimate the Doppler shifts in underwater acoustic OFDM mobile communication for Unmanned Underwater Vehicle.
To overcome disadvantages of Orthogonal Frequency Division Multiplexing(OFDM) underwater acoustic mobile communication, a data variance-based Doppler frequency shift estimation method is proposed to estimate the time-varying Doppler shifts. The estimated channel response of previous OFDM symbol is utilized to recover the valid data sequence and its frequency-domain diversity. The variance of the ratio of data sequence and diversity copy are calculated under different Doppler compensation factors. Then the correct Doppler frequency shift factor is achieved by seeking the minimum of the variance. Sparse Bayesian learning and decision feedback channel estimation algorithm are used for calculating the revised channel frequency-domain response. The channel response is propagated to the follow-up symbol to track the time-varying Doppler shifts. The feasibility and superiority of the proposed method are verified by simulation. The sea trail demonstrates that the proposed algorithm can effectively estimate the Doppler shifts in underwater acoustic OFDM mobile communication for Unmanned Underwater Vehicle.
2022, 44(6): 2045-2053.
doi: 10.11999/JEIT211322
Abstract:
To meet the demand of covert Underwater Acoustic Communication (UAC), a bionic UAC method by mimicking dolphin whistle based on Frequency Shift Keying (FSK) is proposed. The information modulated baseband signal is added to the selected spectrum contour of the dolphin whistle with a certain weight to obtain the synthetic contour. Then generate the synthetic whistle to transmit the information. The receiver extracts the received synthetic whistle, and coherently multiples it with the local whistle, whose contour has a fix frequency difference from the selected whistle contour. A low-pass filtering is adopted to obtain the FSK signal, which is used for demodulation. The bionic effect is analyzed through Time-Frequency Correlation Coefficient (TFCC) and Mel frequency cepstrum Distance (MelD). Simulation and sea trial verified its feasibility. A reliable communication can be achieved at 2km when the symbol width is 0.1s and the TFCC is over 0.99. The low complexity makes the proposed bionic UAC method more suitable for implement, which provides technical support for the practical application of bionic UAC.
To meet the demand of covert Underwater Acoustic Communication (UAC), a bionic UAC method by mimicking dolphin whistle based on Frequency Shift Keying (FSK) is proposed. The information modulated baseband signal is added to the selected spectrum contour of the dolphin whistle with a certain weight to obtain the synthetic contour. Then generate the synthetic whistle to transmit the information. The receiver extracts the received synthetic whistle, and coherently multiples it with the local whistle, whose contour has a fix frequency difference from the selected whistle contour. A low-pass filtering is adopted to obtain the FSK signal, which is used for demodulation. The bionic effect is analyzed through Time-Frequency Correlation Coefficient (TFCC) and Mel frequency cepstrum Distance (MelD). Simulation and sea trial verified its feasibility. A reliable communication can be achieved at 2km when the symbol width is 0.1s and the TFCC is over 0.99. The low complexity makes the proposed bionic UAC method more suitable for implement, which provides technical support for the practical application of bionic UAC.
2022, 44(6): 2054-2060.
doi: 10.11999/JEIT211363
Abstract:
In order to solve the problems of low time synchronization and positioning accuracy for Mobile Underwater Acoustic Asynchronous Networks(MUAANs)caused by nodes movement and long communication time delay, a one-way prediction based joint self-localization and synchronization algorithm is proposed. In this algorithm, each mobile node’s position and time synchronization can be predicted simultaneously by constructing a joint state model and observation model. Meanwhile, the one-way communication is used between each node in the network which can solve the problem of changed communication time delay when nodes are moving at different time and improve the accuracy and time efficiency of self-localization and time synchronization. Simulation results show that the proposed algorithm can predict the mobile node’s position and time synchronization jointly and increase the accuracy and time efficiency of self-localization and time synchronization simultaneously which is more suitable for underwater environment.
In order to solve the problems of low time synchronization and positioning accuracy for Mobile Underwater Acoustic Asynchronous Networks(MUAANs)caused by nodes movement and long communication time delay, a one-way prediction based joint self-localization and synchronization algorithm is proposed. In this algorithm, each mobile node’s position and time synchronization can be predicted simultaneously by constructing a joint state model and observation model. Meanwhile, the one-way communication is used between each node in the network which can solve the problem of changed communication time delay when nodes are moving at different time and improve the accuracy and time efficiency of self-localization and time synchronization. Simulation results show that the proposed algorithm can predict the mobile node’s position and time synchronization jointly and increase the accuracy and time efficiency of self-localization and time synchronization simultaneously which is more suitable for underwater environment.
2022, 44(6): 2061-2070.
doi: 10.11999/JEIT211418
Abstract:
Underwater source passive ranging is based on the pressure radiated by the source in the received data. It is a parameter estimation problem to search for source position parameters in the airspace through the method. Parameter estimation problems are usually converted into classification problems by machine learning methods, which have more accurate estimation capabilities than traditional Matched Field Processing (MFP) and with needless prior sound field information. However, when the probability density of training data and test data follow different distributions or the training data is insufficient, the effect of the classifier under traditional machine learning methods is usually poor. Therefore, an underwater target source ranging algorithm based on Joint Distribution Adaptation (JDA) is proposed to find an appropriate transformation matrix for data mapping, thereby reducing the distribution differences and realizing the migration between source and target. The experimental results indicate that JDA can effectively reduce the differences between the track data obtained in the underwater acoustic field at different times and orientations, thus target could be predicted by classifier based on the source training. The resulting Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are reduced by more than 30%, enabling more accurate distance estimates.
Underwater source passive ranging is based on the pressure radiated by the source in the received data. It is a parameter estimation problem to search for source position parameters in the airspace through the method. Parameter estimation problems are usually converted into classification problems by machine learning methods, which have more accurate estimation capabilities than traditional Matched Field Processing (MFP) and with needless prior sound field information. However, when the probability density of training data and test data follow different distributions or the training data is insufficient, the effect of the classifier under traditional machine learning methods is usually poor. Therefore, an underwater target source ranging algorithm based on Joint Distribution Adaptation (JDA) is proposed to find an appropriate transformation matrix for data mapping, thereby reducing the distribution differences and realizing the migration between source and target. The experimental results indicate that JDA can effectively reduce the differences between the track data obtained in the underwater acoustic field at different times and orientations, thus target could be predicted by classifier based on the source training. The resulting Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are reduced by more than 30%, enabling more accurate distance estimates.
2022, 44(6): 2071-2079.
doi: 10.11999/JEIT220017
Abstract:
Compressive beamforming based on the minimax-concave penalty function constraint, compared with the traditional\begin{document}$ {l_1} $\end{document} ![]()
![]()
Compressive beamforming based on the minimax-concave penalty function constraint, compared with the traditional
2022, 44(6): 2080-2092.
doi: 10.11999/JEIT210376
Abstract:
Associative memory is an important mechanism describing biological learning process and forgetting process, which is of great significance for constructing neuromorphic computing systems, as well as simulating brain-like functions. As a result, the design and implementation of associative memory circuit has become a research hotspot in the field of artificial neural networks. Pavlov conditioning experiment, as one of the classic cases of associative memory, its hardware implementation still suffers from some limitations such as complex circuit configuration, imperfect function and unclear process description. Based on this, a memory circuit is proposed based on memristor full-fuction pavlov associative in this paper, which combines the classical conditioned reflection theory and nano science and technology. Firstly, the Ag/TiOx nanobelt/Ti memristor is prepared using hydrothermal synthesis method and magnetron sputtering method, and its performance testing is conducted jointly by electrochemical workstation, four-probe test bench, and transmission electron microscope. Then, the mathematical model and SPICE circuit model of the Ag/TiOx nanobelt/Ti memristor are built up respectively, based on the electrochemical data derived from the performance testing, and the model accuracy is verified by objective evaluation. Furthermore, the proposed Ag/TiOx nanobelt/Ti memristor model is applied to the implementation of a full-function Pavlovian associative memory circuit. The specific circuit description and function analysis illustrate that this circuit is able to simulate two kinds of learning process and three kinds of forgetting process in Pavlov experiment. Finally, a series of computer simulation and analysis are carried out, which verifies the validity and effectiveness of the entire scheme.
Associative memory is an important mechanism describing biological learning process and forgetting process, which is of great significance for constructing neuromorphic computing systems, as well as simulating brain-like functions. As a result, the design and implementation of associative memory circuit has become a research hotspot in the field of artificial neural networks. Pavlov conditioning experiment, as one of the classic cases of associative memory, its hardware implementation still suffers from some limitations such as complex circuit configuration, imperfect function and unclear process description. Based on this, a memory circuit is proposed based on memristor full-fuction pavlov associative in this paper, which combines the classical conditioned reflection theory and nano science and technology. Firstly, the Ag/TiOx nanobelt/Ti memristor is prepared using hydrothermal synthesis method and magnetron sputtering method, and its performance testing is conducted jointly by electrochemical workstation, four-probe test bench, and transmission electron microscope. Then, the mathematical model and SPICE circuit model of the Ag/TiOx nanobelt/Ti memristor are built up respectively, based on the electrochemical data derived from the performance testing, and the model accuracy is verified by objective evaluation. Furthermore, the proposed Ag/TiOx nanobelt/Ti memristor model is applied to the implementation of a full-function Pavlovian associative memory circuit. The specific circuit description and function analysis illustrate that this circuit is able to simulate two kinds of learning process and three kinds of forgetting process in Pavlov experiment. Finally, a series of computer simulation and analysis are carried out, which verifies the validity and effectiveness of the entire scheme.
2022, 44(6): 2093-2100.
doi: 10.11999/JEIT210328
Abstract:
As the source of the cipher security, the true random number plays an important and irreplaceable role in kinds of cipher systems. In practice, the true random number is usually generated in the random physic processes, and it is vulnerable to the malicious attack, caused by the environment. Thus the poorer quality of the random number makes the cryptographic application security under threat. In order to research the Ring Oscillator-based True Random Number Generator(RO-TRNG), which is widely used currently, the impact on its entropy source and extract is further analyzed and researched, as well as the resulting security variation, in the fault injection attack scenario. Based on the acquisition bias degree, the RO-TRNG security metric model in the scenario of the fault disturbance, is prompted. It is experimented that, while exposed to the fault disturbance environment, the impact on the accumulation of the vibration variance and the bias on the acquistion probability of the rising edge of the beat delay chain for RO-TRNG, can be fully reflected by the model, and the RO-TRNG security can be analyzed objectively, while exposed to the highly dangerous and complicated environment.
As the source of the cipher security, the true random number plays an important and irreplaceable role in kinds of cipher systems. In practice, the true random number is usually generated in the random physic processes, and it is vulnerable to the malicious attack, caused by the environment. Thus the poorer quality of the random number makes the cryptographic application security under threat. In order to research the Ring Oscillator-based True Random Number Generator(RO-TRNG), which is widely used currently, the impact on its entropy source and extract is further analyzed and researched, as well as the resulting security variation, in the fault injection attack scenario. Based on the acquisition bias degree, the RO-TRNG security metric model in the scenario of the fault disturbance, is prompted. It is experimented that, while exposed to the fault disturbance environment, the impact on the accumulation of the vibration variance and the bias on the acquistion probability of the rising edge of the beat delay chain for RO-TRNG, can be fully reflected by the model, and the RO-TRNG security can be analyzed objectively, while exposed to the highly dangerous and complicated environment.
2022, 44(6): 2101-2109.
doi: 10.11999/JEIT210307
Abstract:
The pinched hysteresis loop is the key basis to judge whether a physical device or a mathematical model is a memristor, and its symmetry property is also one of the important characteristics of a memristor. In this paper, an active asymmetric memristive diode-bridge emulator is proposed, whose asymmetry can be controlled by changing the number of parallel diodes in the diode-bridge. Firstly, the fingerprint of this asymmetric memristor emulator is tested, and the effects of excitation frequency and symmetry control parameter on the asymmetry of the pinched hysteresis loop are discussed. Thereafter, by coupling the asymmetric memristor into a Sallen-Key high-pass filter, an inductor-free memristive Chua’s circuit is constructed. The corresponding dimensionless system is built, upon which the asymmetric evolution feature of system attractor is uncovered. Based on the equilibrium stability analysis, bifurcation analysis and multiple attractors distribution in state initial space, the mechanism of attractor asymmetric evolution is clarified. The results demonstrated that, affected by the asymmetric memristor, two unstable saddle-foci of the inductor-free memristive Chua’s circuit are out of balance, resulting in the generations of asymmetric coexistence bifurcation and multi-stable mode. Finally, the correctness of theoretical analysis and numerical simulation are verified by hardware circuit experiment.
The pinched hysteresis loop is the key basis to judge whether a physical device or a mathematical model is a memristor, and its symmetry property is also one of the important characteristics of a memristor. In this paper, an active asymmetric memristive diode-bridge emulator is proposed, whose asymmetry can be controlled by changing the number of parallel diodes in the diode-bridge. Firstly, the fingerprint of this asymmetric memristor emulator is tested, and the effects of excitation frequency and symmetry control parameter on the asymmetry of the pinched hysteresis loop are discussed. Thereafter, by coupling the asymmetric memristor into a Sallen-Key high-pass filter, an inductor-free memristive Chua’s circuit is constructed. The corresponding dimensionless system is built, upon which the asymmetric evolution feature of system attractor is uncovered. Based on the equilibrium stability analysis, bifurcation analysis and multiple attractors distribution in state initial space, the mechanism of attractor asymmetric evolution is clarified. The results demonstrated that, affected by the asymmetric memristor, two unstable saddle-foci of the inductor-free memristive Chua’s circuit are out of balance, resulting in the generations of asymmetric coexistence bifurcation and multi-stable mode. Finally, the correctness of theoretical analysis and numerical simulation are verified by hardware circuit experiment.
2022, 44(6): 2110-2118.
doi: 10.11999/JEIT210216
Abstract:
The design of DNA molecular logic circuits is an important direction in the field of DNA computing. Considering the problems of high complexity and slow response time for dual rail molecular logic circuits, a new strategy based on DNA domain coding is proposed in this study, which is used to construct molecular logic circuits. In this paper, the operation modules of “multiple-inputs-one-output” are introduced, and the fan-out gates and amplification gates are also constructed. Then, the molecular logic circuit to solve four-bits-square-rooting is formed with these logic computing modules designed in this paper. Compared with the four-bit square root circuit under the classical dual-track strategy, the number of reactants is reduced from 130 to 61, and the system response time is reduced to 1 / 24 of the dual-track strategy, which simplifies greatly the complexity of the circuit and improves the response speed of the system. It verifies further the effectiveness of the domain coding strategy in the design of molecular logic circuits. In order to analyze further the design concept for large-scale complicated molecular logic circuits based on domain coding, a four-bit excess-3 code subtracter is constructed, which provides more solutions for designing large-scale functional DNA logic circuits.
The design of DNA molecular logic circuits is an important direction in the field of DNA computing. Considering the problems of high complexity and slow response time for dual rail molecular logic circuits, a new strategy based on DNA domain coding is proposed in this study, which is used to construct molecular logic circuits. In this paper, the operation modules of “multiple-inputs-one-output” are introduced, and the fan-out gates and amplification gates are also constructed. Then, the molecular logic circuit to solve four-bits-square-rooting is formed with these logic computing modules designed in this paper. Compared with the four-bit square root circuit under the classical dual-track strategy, the number of reactants is reduced from 130 to 61, and the system response time is reduced to 1 / 24 of the dual-track strategy, which simplifies greatly the complexity of the circuit and improves the response speed of the system. It verifies further the effectiveness of the domain coding strategy in the design of molecular logic circuits. In order to analyze further the design concept for large-scale complicated molecular logic circuits based on domain coding, a four-bit excess-3 code subtracter is constructed, which provides more solutions for designing large-scale functional DNA logic circuits.
2022, 44(6): 2119-2126.
doi: 10.11999/JEIT210382
Abstract:
Sea clutter and small targets have complex characteristics in the high-resolution radar system. For the target with small radar cross section, the traditional detection method has limited detection performance. In order to break through the critical signal to clutter ratio state, one or more features of radar echo can be extracted for joint feature detection, which is an important way to achieve effective detection in the case of critical signal to clutter ratio. At present, convex hull learning algorithm can be used to calculate the decision region and control effectively the false alarm probability in the feature space of three dimensions and below, but the computational complexity of convex hull learning algorithm is increased above the feature space, and make it difficult to detect target. To solve this problem, a small target detection method based on label propagation algorithm is proposed. It can detect small target in high-dimensional feature space and the false alarm can be effectively controlled. The experimental results on the actual database show, the detection probabilities of 88.4% and 92.0% are obtained in 0.512 s and 1.024 s respectively, which are 3.3% and 2.8% higher than those of the K-Nearest Neighbor (KNN) detector.
Sea clutter and small targets have complex characteristics in the high-resolution radar system. For the target with small radar cross section, the traditional detection method has limited detection performance. In order to break through the critical signal to clutter ratio state, one or more features of radar echo can be extracted for joint feature detection, which is an important way to achieve effective detection in the case of critical signal to clutter ratio. At present, convex hull learning algorithm can be used to calculate the decision region and control effectively the false alarm probability in the feature space of three dimensions and below, but the computational complexity of convex hull learning algorithm is increased above the feature space, and make it difficult to detect target. To solve this problem, a small target detection method based on label propagation algorithm is proposed. It can detect small target in high-dimensional feature space and the false alarm can be effectively controlled. The experimental results on the actual database show, the detection probabilities of 88.4% and 92.0% are obtained in 0.512 s and 1.024 s respectively, which are 3.3% and 2.8% higher than those of the K-Nearest Neighbor (KNN) detector.
2022, 44(6): 2127-2134.
doi: 10.11999/JEIT210335
Abstract:
Synthetic Aperture Radar (SAR) imaging has a large amount of data volume, high sampling rate, and the problem of SAR imaging precision in traditional Compression Sensing (CS) is low, and there is a problem of poor anti-noise performance. A method of reconstruction method of two - dimensional sampling synthetic aperture rada based on Iterative Proximal Projection (IPP) is proposed. The radar echo is constructed as a two-dimensional sparse representation model in the range frequency-domain-azimuth Doppler region. On this basis, the two-dimensional imaging problem is transformed into the range and azimuth compression sensing sparse reconstruction. The function optimization model of the iterative proximal projection algorithm is used to represent the sparse representation of the synthetic aperture thunder imaging, and the proximal operator is finally obtained with the Smoothly Clipped Absolute Deviation (SCAD) penalty function to solve the model and to image. Simulation and measured data processing results show that the method of imaging is better.
Synthetic Aperture Radar (SAR) imaging has a large amount of data volume, high sampling rate, and the problem of SAR imaging precision in traditional Compression Sensing (CS) is low, and there is a problem of poor anti-noise performance. A method of reconstruction method of two - dimensional sampling synthetic aperture rada based on Iterative Proximal Projection (IPP) is proposed. The radar echo is constructed as a two-dimensional sparse representation model in the range frequency-domain-azimuth Doppler region. On this basis, the two-dimensional imaging problem is transformed into the range and azimuth compression sensing sparse reconstruction. The function optimization model of the iterative proximal projection algorithm is used to represent the sparse representation of the synthetic aperture thunder imaging, and the proximal operator is finally obtained with the Smoothly Clipped Absolute Deviation (SCAD) penalty function to solve the model and to image. Simulation and measured data processing results show that the method of imaging is better.
2022, 44(6): 2135-2142.
doi: 10.11999/JEIT210354
Abstract:
Retracking is an important step to extract accurate parameter estimation of the altimeter echo signals. The existing retracking estimators of Synthetic Aperture Radar (SAR) altimeter are mainly based on the least square method, which does not consider the influence of the different statistical characteristics of each range bin of the altimeter on the parameter estimation accuracy. At the same time, the existing weighted least squares retracking estimator uses a semi-analytical model, which is inefficient and inconvenient be applied to practice. In this paper, a new kind of weighted method is designed by using the analytical SAR Altimetry MOde Studies and Applications (SAMOSA) echo model for data processing. The weighting process makes the statistical characteristics of each range bin of the altimeter consistent, thereby improving the estimation accuracy of the range and significant wave height of the retracking estimator. The method is verified by Sentinel-3A satellite data, and the verification results show that the estimation accuracy of the two parameters, range and significant wave height, are improved: at 2 m significant wave height, the range accuracy is improved by 9%, and the significant wave height accuracy is improved by 11%; at 4 m significant wave height, the range accuracy is improved by 13%, and the significant wave height accuracy is improved by 20%.
Retracking is an important step to extract accurate parameter estimation of the altimeter echo signals. The existing retracking estimators of Synthetic Aperture Radar (SAR) altimeter are mainly based on the least square method, which does not consider the influence of the different statistical characteristics of each range bin of the altimeter on the parameter estimation accuracy. At the same time, the existing weighted least squares retracking estimator uses a semi-analytical model, which is inefficient and inconvenient be applied to practice. In this paper, a new kind of weighted method is designed by using the analytical SAR Altimetry MOde Studies and Applications (SAMOSA) echo model for data processing. The weighting process makes the statistical characteristics of each range bin of the altimeter consistent, thereby improving the estimation accuracy of the range and significant wave height of the retracking estimator. The method is verified by Sentinel-3A satellite data, and the verification results show that the estimation accuracy of the two parameters, range and significant wave height, are improved: at 2 m significant wave height, the range accuracy is improved by 9%, and the significant wave height accuracy is improved by 11%; at 4 m significant wave height, the range accuracy is improved by 13%, and the significant wave height accuracy is improved by 20%.
2022, 44(6): 2143-2150.
doi: 10.11999/JEIT210397
Abstract:
The separation between the Periodic Frequency Modulation (PFM) interference and satellite navigation signal is not enough in the traditional transform domain, which causes severe satellite navigation signal degradation is suppressing interference. To solve this problem, a data rearrangement method based on period truncation is proposed. Using the periodicity of the PFM interfering signal, the energy scattered in a larger bandwidth is concentrated to a single frequency point in the rearranged data. Then, Singular Value Decomposition (SVD) is adopted to map the interference and the desired signal into different projection subspaces to eliminate the interference components. The simulation results show that the proposed method can reduce the overlap degree of the PFM interference and satellite navigation signals, and reduce the damage of satellite navigation signals.
The separation between the Periodic Frequency Modulation (PFM) interference and satellite navigation signal is not enough in the traditional transform domain, which causes severe satellite navigation signal degradation is suppressing interference. To solve this problem, a data rearrangement method based on period truncation is proposed. Using the periodicity of the PFM interfering signal, the energy scattered in a larger bandwidth is concentrated to a single frequency point in the rearranged data. Then, Singular Value Decomposition (SVD) is adopted to map the interference and the desired signal into different projection subspaces to eliminate the interference components. The simulation results show that the proposed method can reduce the overlap degree of the PFM interference and satellite navigation signals, and reduce the damage of satellite navigation signals.
2022, 44(6): 2151-2157.
doi: 10.11999/JEIT210274
Abstract:
As a new type of low-frequency magnetic antenna, the rotating magnet antenna overcomes the shortcomings including large size, high power consumption, and low radiation efficiency of traditional low-frequency antenna and is mainly used in near-field low-frequency magnetic communication. FEKO is used to analyze the change of the magnetic field in the near area of the antenna with the direction, and explore the influence of the infinite ground on the magnetic field distribution in the near area of the antenna. The result shows that the antenna has the largest radiation intensity in its radial direction and the smallest radiation intensity in its axial direction. And the earth’s surface has less influence on the magnetic field signal in the short communication distance range. The relationship between magnetic and moment of inertia is analyzed, and the dimension parameters are optimized. An optimized prototype antenna and an original antenna are manufactured, and the power consumption of the two antennas during operating are measured. The experimental results demonstrate that the optimized antenna, which weighs 30 g more than the original antenna, reduces power consumption of about 5.5 W. The direct antenna modulation method is used to 2FSK modulation of the near-zone magnetic field, and the original symbol information is recovered using non-coherent demodulation. The measured results show that the optimized antenna can achieve ultra-low frequency communication with a symbol rate of 3.5 bps within a 20 m range in harsh electromagnetic environment.
As a new type of low-frequency magnetic antenna, the rotating magnet antenna overcomes the shortcomings including large size, high power consumption, and low radiation efficiency of traditional low-frequency antenna and is mainly used in near-field low-frequency magnetic communication. FEKO is used to analyze the change of the magnetic field in the near area of the antenna with the direction, and explore the influence of the infinite ground on the magnetic field distribution in the near area of the antenna. The result shows that the antenna has the largest radiation intensity in its radial direction and the smallest radiation intensity in its axial direction. And the earth’s surface has less influence on the magnetic field signal in the short communication distance range. The relationship between magnetic and moment of inertia is analyzed, and the dimension parameters are optimized. An optimized prototype antenna and an original antenna are manufactured, and the power consumption of the two antennas during operating are measured. The experimental results demonstrate that the optimized antenna, which weighs 30 g more than the original antenna, reduces power consumption of about 5.5 W. The direct antenna modulation method is used to 2FSK modulation of the near-zone magnetic field, and the original symbol information is recovered using non-coherent demodulation. The measured results show that the optimized antenna can achieve ultra-low frequency communication with a symbol rate of 3.5 bps within a 20 m range in harsh electromagnetic environment.
2022, 44(6): 2158-2165.
doi: 10.11999/JEIT210349
Abstract:
In order to solve the optimization problem of jamming resource scheduling in medium and large-scale Unmanned Aerial Vehicle (UAV) jamming scenarios, a jamming resource scheduling model that can meet the minimum number of tasks constraint is proposed to improve the simple constraints and small-scale solution algorithms of the existing models. The interference benefit and cost indicators are weighted by the analytic hierarchy process. Then an improved parallel genetic algorithm is designed, where the elite set is introduced to accelerate the convergence of the algorithm. The simulation results in medium scale and larger scale jamming situations, such as 500:500 (number of jamming resources: number of targets) show that the proposed algorithm converges faster and achieves better objective function value than the existing representative and improved genetic algorithms.
In order to solve the optimization problem of jamming resource scheduling in medium and large-scale Unmanned Aerial Vehicle (UAV) jamming scenarios, a jamming resource scheduling model that can meet the minimum number of tasks constraint is proposed to improve the simple constraints and small-scale solution algorithms of the existing models. The interference benefit and cost indicators are weighted by the analytic hierarchy process. Then an improved parallel genetic algorithm is designed, where the elite set is introduced to accelerate the convergence of the algorithm. The simulation results in medium scale and larger scale jamming situations, such as 500:500 (number of jamming resources: number of targets) show that the proposed algorithm converges faster and achieves better objective function value than the existing representative and improved genetic algorithms.
2022, 44(6): 2166-2174.
doi: 10.11999/JEIT210386
Abstract:
To address the shortcomings of existing low illumination image enhancement algorithms in achieving detail enhancement while considering noise suppression, a reference-free low-illumination image enhancement method based on deep convolutional neural networks is proposed in the paper. First, the illumination and reflection components are extracted from the input low-illumination image based on Retinex theory and optimised separately, after which the optimised illumination and reflection components are multiplied to obtain the enhanced image. loss to update the network parameters, meanwhile, the denoising effect of Block Matching 3D (BM3D ) is integrated into the optimization process of reflection components. The experimental results show that the algorithm in this paper can effectively enhance the contrast and brightness of low-illumination images compared to existing mainstream algorithms, while maintaining the naturalness of the images.
To address the shortcomings of existing low illumination image enhancement algorithms in achieving detail enhancement while considering noise suppression, a reference-free low-illumination image enhancement method based on deep convolutional neural networks is proposed in the paper. First, the illumination and reflection components are extracted from the input low-illumination image based on Retinex theory and optimised separately, after which the optimised illumination and reflection components are multiplied to obtain the enhanced image. loss to update the network parameters, meanwhile, the denoising effect of Block Matching 3D (BM3D ) is integrated into the optimization process of reflection components. The experimental results show that the algorithm in this paper can effectively enhance the contrast and brightness of low-illumination images compared to existing mainstream algorithms, while maintaining the naturalness of the images.
2022, 44(6): 2175-2183.
doi: 10.11999/JEIT210344
Abstract:
Focusing on the problem of low utilization of object features and inaccurate detection results in CenterNet, an improved algorithm of double branch feature fusion is proposed in the paper. One branch of the algorithm includes feature pyramid enhancement module and feature fusion module to fuse the multi-layer features output from the backbone network. At the same time, in order to use more high-level semantic information, only the last layer of the backbone network is upsampled in the other branch. Secondly, a frequency-based channel attention mechanism is added to the backbone network to enhance feature extraction capability. Finally, the features of the two branches are concatenated and convoluted. The experimental results show that the detection accuracy on PASCAL VOC dataset is 82.3%, which is 3.6% higher than CenterNet, and the detection accuracy on KITTI dataset is 6% higher than CenterNet. The detection speed meets the real-time requirements. The double branch feature fusion method is proposed to process the features of different layers, which makes better use of the spatial information of shallow features and the semantic information of deep features, and improves the detection performance of the algorithm.
Focusing on the problem of low utilization of object features and inaccurate detection results in CenterNet, an improved algorithm of double branch feature fusion is proposed in the paper. One branch of the algorithm includes feature pyramid enhancement module and feature fusion module to fuse the multi-layer features output from the backbone network. At the same time, in order to use more high-level semantic information, only the last layer of the backbone network is upsampled in the other branch. Secondly, a frequency-based channel attention mechanism is added to the backbone network to enhance feature extraction capability. Finally, the features of the two branches are concatenated and convoluted. The experimental results show that the detection accuracy on PASCAL VOC dataset is 82.3%, which is 3.6% higher than CenterNet, and the detection accuracy on KITTI dataset is 6% higher than CenterNet. The detection speed meets the real-time requirements. The double branch feature fusion method is proposed to process the features of different layers, which makes better use of the spatial information of shallow features and the semantic information of deep features, and improves the detection performance of the algorithm.
2022, 44(6): 2184-2194.
doi: 10.11999/JEIT210321
Abstract:
Most existing knowledge representation learning models treat knowledge triples independently, it fail to cover and leverage the feature information in any given entity’s neighborhood. Besides, embedding knowledge graphs with tree-like hierarchical structure in Euclidean space would incur a large distortion in embeddings. To tackle such issues, a link prediction method based on Hyperbolic Graph ATtention networks for Link Prediction in knowledge graphs (HyGAT-LP) is proposed. Firstly, knowledge graphs are embedded in hyperbolic space with constant negative curvature, which is more suited for knowledge graphs’ tree-like hierarchical structure. Then the proposed method aggregates feature information in the given entity’s neighborhood with both entity-level and relation-level attention mechanisms, and further, embeds the given entity in low dimensional hyperbolic space. Finally, every triple’s score is computed by a scoring function, and links in knowledge graphs are predicted based on the scores indicating the probabilities that predicted triples are correct. Experimental results show that, compared with baseline models, the proposed method can significantly improve the performance of link prediction in knowledge graphs.
Most existing knowledge representation learning models treat knowledge triples independently, it fail to cover and leverage the feature information in any given entity’s neighborhood. Besides, embedding knowledge graphs with tree-like hierarchical structure in Euclidean space would incur a large distortion in embeddings. To tackle such issues, a link prediction method based on Hyperbolic Graph ATtention networks for Link Prediction in knowledge graphs (HyGAT-LP) is proposed. Firstly, knowledge graphs are embedded in hyperbolic space with constant negative curvature, which is more suited for knowledge graphs’ tree-like hierarchical structure. Then the proposed method aggregates feature information in the given entity’s neighborhood with both entity-level and relation-level attention mechanisms, and further, embeds the given entity in low dimensional hyperbolic space. Finally, every triple’s score is computed by a scoring function, and links in knowledge graphs are predicted based on the scores indicating the probabilities that predicted triples are correct. Experimental results show that, compared with baseline models, the proposed method can significantly improve the performance of link prediction in knowledge graphs.
2022, 44(6): 2195-2206.
doi: 10.11999/JEIT210322
Abstract:
The population size is the most significant parameter to determine the performance of the algorithm, and its size may cause problems such as premature convergence or low efficiency of the algorithm. A dynamic control method of Population Size besed on Euclidean Distance (EDPS) is proposed. The core circle is established by adopting the Euclidean distance, and the feedback information of the core circle is used to control dynamically the population size, and the method of increasing or deleting the number of individuals based on the core circle is proposed. The strategy is applied to particle swarm optimization algorithm, genetic algorithm and differential evolution algorithm, whose performance is verified in the test functions. The experimental results show the proposed new strategy is effective.
The population size is the most significant parameter to determine the performance of the algorithm, and its size may cause problems such as premature convergence or low efficiency of the algorithm. A dynamic control method of Population Size besed on Euclidean Distance (EDPS) is proposed. The core circle is established by adopting the Euclidean distance, and the feedback information of the core circle is used to control dynamically the population size, and the method of increasing or deleting the number of individuals based on the core circle is proposed. The strategy is applied to particle swarm optimization algorithm, genetic algorithm and differential evolution algorithm, whose performance is verified in the test functions. The experimental results show the proposed new strategy is effective.
2022, 44(6): 2207-2215.
doi: 10.11999/JEIT210333
Abstract:
The low spatial resolution characteristics of hyperspectral images often make it difficult for global texture extraction techniques to obtain accurate texture information and the single-scale local texture extraction technology is not satisfactory for effectively identifying the features. In this article, a Multi-scale Superpixel Texture Preservation and Fusion is proposed for hyperspectral image classification. Specifically, the original hyperspectral image is first extracted with multi-direction and scale global texture using 2D Gabor filter, and the texture feature of each scale is merged to enhance the texture structure characterization ability. Next, texture and spectral principal component features are fused to form spectral-texture joint discriminant features. After that, the shape adaptive oversegmentation method is applied to the spectral-texture joint feature for local texture information preservation and fusion. In particular, in order to overcome the hidden irrelevance problem of neighboring pixels, a density-based nearest neighbor similarity evaluation criterion is defined, which aims to make the superpixel texture more consistent. Finally, the updated spectral-texture joint discriminant features are input into the pixel-level classifiers to obtain their corresponding class labels, and the decision fusion mechanism of majority voting is adopted to obtain the final classification result. Experiments on the real data sets of Indian Pines and Pavia University show that the classification accuracy of this method under the condition of small samples is better than eight comparison methods such as the benchmark classifier Support Vector Machine (SVM), deep learning method Gabor Filtering and Deep Network (GFDN), and the latest spatial-spectral method Spectral-Spatial and Superpixelwise Principal Component Analysis (S3-PCA), which proves fully the practicability and effectiveness of the proposed method.
The low spatial resolution characteristics of hyperspectral images often make it difficult for global texture extraction techniques to obtain accurate texture information and the single-scale local texture extraction technology is not satisfactory for effectively identifying the features. In this article, a Multi-scale Superpixel Texture Preservation and Fusion is proposed for hyperspectral image classification. Specifically, the original hyperspectral image is first extracted with multi-direction and scale global texture using 2D Gabor filter, and the texture feature of each scale is merged to enhance the texture structure characterization ability. Next, texture and spectral principal component features are fused to form spectral-texture joint discriminant features. After that, the shape adaptive oversegmentation method is applied to the spectral-texture joint feature for local texture information preservation and fusion. In particular, in order to overcome the hidden irrelevance problem of neighboring pixels, a density-based nearest neighbor similarity evaluation criterion is defined, which aims to make the superpixel texture more consistent. Finally, the updated spectral-texture joint discriminant features are input into the pixel-level classifiers to obtain their corresponding class labels, and the decision fusion mechanism of majority voting is adopted to obtain the final classification result. Experiments on the real data sets of Indian Pines and Pavia University show that the classification accuracy of this method under the condition of small samples is better than eight comparison methods such as the benchmark classifier Support Vector Machine (SVM), deep learning method Gabor Filtering and Deep Network (GFDN), and the latest spatial-spectral method Spectral-Spatial and Superpixelwise Principal Component Analysis (S3-PCA), which proves fully the practicability and effectiveness of the proposed method.
2022, 44(6): 2216-2229.
doi: 10.11999/JEIT210161
Abstract:
Specific radar emitter identification distinguishes each radar emitter based on the extracted individual features, which is crucial for electronic countermeasures. With the rapid development of deep learning, specific radar emitter identification using deep learning architecture draws great attention recently. Despite many years of research and rich achievements, there is still lack of a comprehensive review about specific radar emitter identification at present. Therefore, a systematic review is provided in this paper from four aspects: (1) the mechanism analysis of identification; (2) the handcrafted feature-based identification methods; (3) the deep learning-based identification methods; (4) and the testing datasets. Finally, the current status and the future directions are summarized, aiming at promoting the new development of specific radar emitter identification.
Specific radar emitter identification distinguishes each radar emitter based on the extracted individual features, which is crucial for electronic countermeasures. With the rapid development of deep learning, specific radar emitter identification using deep learning architecture draws great attention recently. Despite many years of research and rich achievements, there is still lack of a comprehensive review about specific radar emitter identification at present. Therefore, a systematic review is provided in this paper from four aspects: (1) the mechanism analysis of identification; (2) the handcrafted feature-based identification methods; (3) the deep learning-based identification methods; (4) and the testing datasets. Finally, the current status and the future directions are summarized, aiming at promoting the new development of specific radar emitter identification.
2022, 44(6): 2230-2244.
doi: 10.11999/JEIT210330
Abstract:
SRAM-based FPGAs are sensitive to single event effect in space radiation environment, resulting in soft errors. Triple Modular Redundancy (TMR) is the most widely used circuit hardening technology to alleviate FPGA soft errors. This paper introduces first the current research status of TMR technology, and then summarizes four key technologies and their implementation principles of fine-grained TMR technology, system partitioning technology, configuration scrubbing technology and state synchronization technology, which are commonly used in TMR tools. As the high-level synthesis technology of FPGA becomes more and more mature, the TMR tools based on high level synthesis have gradually become a new research branch. The current mainstream TMR tools based on the register transfer level, TMR tools based on important soft-core resources, and the emerging TMR tools based on high-level synthesis are classified and introduced. Finally, the future development trend of TMR tool for FPGA is summarized and forecasted.
SRAM-based FPGAs are sensitive to single event effect in space radiation environment, resulting in soft errors. Triple Modular Redundancy (TMR) is the most widely used circuit hardening technology to alleviate FPGA soft errors. This paper introduces first the current research status of TMR technology, and then summarizes four key technologies and their implementation principles of fine-grained TMR technology, system partitioning technology, configuration scrubbing technology and state synchronization technology, which are commonly used in TMR tools. As the high-level synthesis technology of FPGA becomes more and more mature, the TMR tools based on high level synthesis have gradually become a new research branch. The current mainstream TMR tools based on the register transfer level, TMR tools based on important soft-core resources, and the emerging TMR tools based on high-level synthesis are classified and introduced. Finally, the future development trend of TMR tool for FPGA is summarized and forecasted.
