Search

Hand gesture recognition based on micro‐Doppler (MD) radar has garnered considerable attention from researchers as a potential method for human–computer interaction (HCI). However, two significant challenges, that is, obvious differences in MD map size and lots of redundant information contained in the MD maps, are encountered. Here, a multi‐scale graph‐based hand gesture recognition framework is proposed. First, a MD graph representation method is developed, which is adapted to arbitrary‐sized frames and enables to map the key features in a sparse manner. Then, the multi‐scale information from MD graph is fully extracted for hand gesture recognition. Experimental results show that the proposed framework achieves a state‐of‐the‐art accuracy of 96.73% on the four‐class radar hand gesture dataset, while reducing up to 99% of the redundant information in the MD maps. This framework requires only a little amount of memory storage with good hand gesture recognition capability, demonstrating its high potential in the HCI field. [ABSTRACT FROM AUTHOR]

The accuracy of radar emitter signal sorting nowadays deteriorates due to the high flexibility and complexity of modern radar pulse streams and the density of crowded electromagnetic environment. In modern radar signal sorting based on pulse repetition interval, conventional methods usually fail to achieve acceptable accuracy and lack stable performance for two main reasons: (1) Conventional methods require a large number of pulses in the stream, which is not practical in many applications. (2) These methods are sensitive to pulse loss and random noise pulses. These two reasons are the main problem that is addressed in this paper. Our proposed model is a machine learning architecture called Unified Residual Recurrent Neural Network (URRNN). In this architecture, residual neural network and recurrent neural network are combined and modified to alleviate the forementioned shortcomings of traditional approaches and enhance the model performance in both classification and deinterleaving tasks. This aim is achieved due to the fact that URRNN extracts both spatial and temporal features, which means more information about processed stream that is exploited to enhance model performance. Three different architectural combinations of URRNN, which show high accuracy and reasonable processing time, are built and trained. The structural and functional description are provided for each architecture. Simulation demonstrates high accuracy and reliable performance of the proposed methods in different circumstances. The results are compared with the results obtained by other conventional machine learning techniques. [ABSTRACT FROM AUTHOR]

The existing automatic target recognition (ATR) methods for synthetic aperture radar (SAR) images mainly utilize the real-valued magnitude information while often ignoring the phase information. However, the phase information also provides important details, which can be utilized to improve the ATR performance. To address this issue, a fully complex-valued light-weight network (CVLWNet) is proposed based on complex-valued operations, such as complex-valued convolution and complex-valued batch normalization. Besides, to achieve reduced parameters and enhanced robustness of the designed network, many complex-valued blocks of operations are built, including the CMish activation function, the complex-valued residual link block (CVReLBlock), the lightweight complex-valued cross stage partial block (LC-CSPBlock). In the designed CVLWNet, the input, output, and weight parameters are all complex-valued, which makes it possible to sufficiently exploit the complex-valued characteristics of SAR data. Comparative experiments are conducted with the moving and stationary target acquisition and recognition dataset. Compared with the state-of-the-art real-valued and complex-valued models under both standard and extended operating conditions, the performance of proposed method is verified.

This work introduces an attention mechanism that can be integrated into any standard convolution neural network to improve model sensitivity and prediction accuracy with minimal computational overhead. The attention mechanism is introduced in a lightweight network – Alexnet and its classification performance for human micro‐Doppler signatures is evaluated. The Alexnet model trained with an attention module can implicitly highlight the salient regions in the radar signatures while suppressing the irrelevant background regions and consistently improving network predictions. Network visualizations are provided through class activation mapping, providing better insights into how the predictions are made. The visualizations demonstrate how the attention mechanism focusses on the region of interest in the radar signatures. [ABSTRACT FROM AUTHOR]

Abstract Continuous Human Activity Recognition (HAR) in arbitrary directions is investigated in this paper using a network of five spatially distributed pulsed Ultra‐Wideband radars. While activities performed continuously and in unconstrained trajectories provide a more realistic and natural scenario for HAR, the network of radar sensors is proposed to address the issue of unfavourable or occluded perspectives when using only a single sensor. Different techniques to combine the relevant information from the multiple radars in the network are investigated, focussing on signal level fusion directly applied on Range‐Time maps, and the selection of radar nodes based on location and velocity of the target derived from multilateration processing and tracking. Recurrent Neural Networks with and without bidirectionality are used to classify the activities based on the micro‐Doppler (μD) spectrograms obtained for sensor fusion techniques. To assess classification performances, novel evaluation metrics accounting for the continuous nature of the sequence of activities and inherent imbalances in the dataset are proposed and compared with existing metrics. It is shown that the conventional accuracy metric may not capture all the important aspects for continuous HAR, and the proposed metrics can be considered for a more comprehensive evaluation.

Radar enables a convenient, contactless way for gait analysis in daily scenarios, such as in‐home health monitoring. However, the overlapping of human limb components could occur in both range and velocity dimensions of the radar echo, which makes it difficult to distinguish the limbs. In this study, a trajectory tracking and recombination method based on range‐Doppler‐time data is proposed for gait feature estimation under insufficient radar resolution. In the trajectory tracking step, the scattering amplitude is included for Kalman filtering, and the weighted joint nearest neighbour algorithm is proposed for data association. Then, the trajectories are recombined to construct the limb tracks based on the relative range‐Doppler angles. Finally, the limb tracks are used to estimate the spatial‐temporal and kinematic gait features. In our simulation, the trajectory tracking accuracies range from 98.2% to 99.6%. The gait feature estimation accuracies for simulation and measurement are no less than 91.9% and 93.8% respectively. [ABSTRACT FROM AUTHOR]

Sparse representation-based methods have shown promising prospects in synthetic aperture radar (SAR) target recognition due to their robustness under complex application conditions. However, these methods are still susceptible to the feature representation of SAR images. In this paper, we propose a new SAR target recognition method using an adaptive kernel sparse representation model based on local contrast perception (LCP). The LCP method combines a specially designed random projection matrix with two-dimensional filtering to capture extensive intensity contrasts between local regions in SAR images. A response map reconstruction technique shows that LCP can obtain discriminative properties between target and background regions and improve feature representation. The LCP features are then transformed into the kernel representation to improve linear separability by defining a reproducing kernel Hilbert space (RKHS) and using a nonlinear kernel function. The kernel representation is utilized to construct a new kernel sparse representation model for classification. Furthermore, neighborhood analysis (NA) is proposed to establish the relationship between the parameter and feature distribution, with an adaptive regularization parameter selection strategy to mitigate the parameter sensitivity. Experiments conducted on the moving and stationary target acquisition and recognition dataset demonstrate the superiority and robustness of the proposed method under different application conditions. • Enhancing SAR target image description with local contrast perception. • Local contrast features promote kernel sparse representation classification. • Learning feature distribution to optimize parameters of sparse representation. • Superior performance under various operating conditions in SAR target recognition. [ABSTRACT FROM AUTHOR]

Abstract In the complex electromagnetic environment of cities, detection and positioning of slow and small unmanned aerial vehicles (UAVs) have always been a very challenging research topic. Usually, multiple radar observation stations are required to detect and track targets. Moreover, each radar observation station requires strict synchronisation, and the system is complex and costly. In view of this, under the condition of known prior information of urban geography, this study proposes a method to locate UAVs in complex electromagnetic environments, which use multiple virtual mirror synchronous radars constructed by multipaths and created by the complex buildings in urban environments. First, the simulation model is constructed, and the corresponding simulation parameters are given. Then, the algorithm for solving the single‐station radar location is proposed, and the algorithm is verified by the simulation of a ray propagation model. Finally, the proposed algorithm is further verified based on the measured data captured by the radar system in Section 2. The results show that the proposed algorithm based on the single‐station radar is able to locate the UAV in both line‐of‐sight and non‐line‐of‐sight site, and its positioning accuracy is better than 1 m.

Abstract Due to the high‐speed rotors and overlapped propellers, it is still a challenge to extract the micro‐motion parameters of the quadcopter. In this work, the authors establish a radar backscattering model of the quadcopter based on the flight control theory and radar backscattering characteristics. Then, the improved template matching method and peak grouping method are proposed to capture the periodic information for overlapped propellers on a range‐time profile and extract the rotation speeds of rotors. By simulating the nine types of quadcopter movements, the relative error of the rotation speeds is estimated to be lower than 1.4% for each rotor with a signal‐to‐noise ratio of 15 dB.

A recognition method combining Cameron decomposition and fusing Reduced Kernel Extreme Learning Machine (RKELM) is proposed for the Full Polarimetric (FP) High Resolution Range Profile (HRRP)-based radar target recognition task. In the feature extraction phase, Cameron decomposition is exploited to define the projection component of the target on the standard scatterers. Through analysis, the projection components on three scattering bases, i.e., trihedral, dihedral, and 1/4 wave device, are selected as target features, which achieve more detailed descriptions of the target characteristics. In the classification phase, considering the instability of the recognition performance of the RKELM algorithm, the RKELM based on prototype clustering preprocessing is first proposed. Then, to improve the recognition performance, we proposed the feature level fusing RKELM and the decision level fusing RKELM to fuse the three projection components of the targets. The experiments compared the performance of the proposed recognition method and the state-of-the-art methods using the FP HRRP data from 10 civilian vehicles. The results demonstrate that the projection features by Cameron decomposition exhibit higher separability and better noise robustness, and that the feature level fusing RKELM has better generalization performance with a large number of training samples, but the decision level fusing RKELM was better with a small number of training samples.

Abstract The radar cross section (RCS) is an important parameter that reflects the scattering characteristics of radar targets. Based on the monostatic radar RCS time series' statistical features by sliding window segmentation, a novel sliding window‐statistical‐gated recurrent unit (SW‐S‐GRU) method for radar target recognition (RTR) is proposed using a GRU. The sliding window method divides the RCS time series into many segments for obtaining the statistical features as the input of the GRU. The deeper features are extracted from the segmented time series and then classified by the GRU. Under the model of precession motion, simulation results show that the proposed method achieves a great recognition accuracy. Even under the condition of a low SNR, the proposed method can obtain better recognition accuracy. For the precession dataset, the effect of sliding window parameters on RTR is analysed. The simulation results show that the recognition result of the model is better using the longer series overlap length between adjacent times. And when the number of hidden layers is 8 or above 8, the recognition accuracy does not change. Using the publicly available measured drone RCS data, the proposed method can also identify the drone well.

Abstract The vortex electromagnetic (EM) wave radar has the potential to obtain more accurate micro‐motion parameters for target recognition. However, with the existing algorithms of micro‐motion parameter extraction it is difficult to obtain the real rotation radius and tilt angle of a rotational target in the presence of multiple scattering points in the radar beam. A micro‐motion parameter extraction algorithm for rotating targets based on the vortex EM wave radar is proposed in this article. The angular Doppler is obtained from the dual‐mode vortex EM echoes. The time interval between the maximum and minimum angular Doppler frequency is derived. The relationship between the time interval and micro‐motion parameters is shown. By combining the linear Doppler and the angular Doppler, the micro‐motion parameters are roughly estimated. Then, fine micro‐motion parameters are obtained by using an iterative soft threshold algorithm. The proposed algorithm can extract the real rotation radius and tilt angle in the case of multiple scattering points. The performance and robustness of the algorithm are proved by simulations.

The target and sea clutter Doppler domains frequently overlap due to the frequent passage of slow ship targets through the sea clutter zone. In this letter, a novel sea clutter suppression method is suggested as a solution to this issue, whose key is a novel singular value zeroing criterion guided by the search results of two‐dimensional spectral peaks. Verified by simulations, the method proposed can improve signal‐to‐clutter ratio (SCR) from −8 to 41 dB in the frequency domain and be more effective than the conventional SVD‐FRFT (a method for joint application of singular value decomposition and fractional Fourier transformer) and improved SVD‐FRFT. [ABSTRACT FROM AUTHOR]

Copyright of Systems Engineering & Electronics is the property of Journal of Systems Engineering & Electronics Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Convolutional Neural Network (CNN) has been widely applied in the field of synthetic aperture radar (SAR) image recognition. Nevertheless, CNN-based recognition methods usually encounter the problem of poor feature representation ability due to insufficient labeled SAR images. In addition, the large inner-class variety and high cross-class similarity of SAR images pose a challenge for classification. To alleviate the problems mentioned above, we propose a novel few-shot learning (FSL) method for SAR image recognition, which is composed of the multi-feature fusion network (MFFN) and the weighted distance classifier (WDC). The MFFN is utilized to extract input images' features, and the WDC outputs the classification results based on these features. The MFFN is constructed by adding a multi-scale feature fusion module (MsFFM) and a hand-crafted feature insertion module (HcFIM) to a standard CNN. The feature extraction and representation capability can be enhanced by inserting the traditional hand-crafted features as auxiliary features. With the aid of information from different scales of features, targets of the same class can be more easily aggregated. The weight generation module in WDC is designed to generate category-specific weights for query images. The WDC distributes these weights along the corresponding Euclidean distance to tackle the high cross-class similarity problem. In addition, weight generation loss is proposed to improve recognition performance by guiding the weight generation module. Experimental results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset and the Vehicle and Aircraft (VA) dataset demonstrate that our proposed method surpasses several typical FSL methods. [ABSTRACT FROM AUTHOR]

For the problem of target recognition in synthetic aperture radar (SAR) image, a method of SAR target recognition based on improved convolution neural network (CNN) and data augmentation is proposed. Firstly, Dropout is brought in the training phase to randomly delete some neurons, so that the generalization ability of the network is enhanced. Secondly, L2 regularization is introduced into the network to reduce the structural risk and effectively restrain the over fitting. Then, Adam is used to optimize the network to improve the convergence efficiency of the model. Finally, the preferred rotation data augmentation method is employed for expanding the data set of SAR target. Through the improved network and increased data, the recognition accuracy and generalization of the model are enhanced. Experiments on moving and stationary target acquisition and recognition (MSTAR) data set show that the proposed method has higher recognition accuracy and better generalization. [ABSTRACT FROM AUTHOR]

Continuous Human Activity Recognition (HAR) in arbitrary directions is investigated in this paper using a network of five spatially distributed pulsed Ultra‐Wideband radars. While activities performed continuously and in unconstrained trajectories provide a more realistic and natural scenario for HAR, the network of radar sensors is proposed to address the issue of unfavourable or occluded perspectives when using only a single sensor. Different techniques to combine the relevant information from the multiple radars in the network are investigated, focussing on signal level fusion directly applied on Range‐Time maps, and the selection of radar nodes based on location and velocity of the target derived from multilateration processing and tracking. Recurrent Neural Networks with and without bidirectionality are used to classify the activities based on the micro‐Doppler (μD) spectrograms obtained for sensor fusion techniques. To assess classification performances, novel evaluation metrics accounting for the continuous nature of the sequence of activities and inherent imbalances in the dataset are proposed and compared with existing metrics. It is shown that the conventional accuracy metric may not capture all the important aspects for continuous HAR, and the proposed metrics can be considered for a more comprehensive evaluation. [ABSTRACT FROM AUTHOR]

In recent years, radar target recognition based on deep learning has received great attention. However, there are time constraints and resource constraints in actual combat, and the problem of small sample recognition greatly iimits its pefformance in actual recognition tasks. To solve this problem, based on the meta learning algorithm, this paper improves the performance of new tasks by learning meta knowledge from multiple related tasks, introduces the idea of tranffer learning, proposes an improved small sample learning method, and analyzes the application boundary conditions of the method through detalled performance comparison experiments. The experimental resuits based on the measured high resolution range proille data show that the pefformance advantage of meta learning method is outstanding only when there are many target categories in the historical accumulated samples and in the case of minimal samples with high correlation with the target task, the proposed method can significantly improve its comprehensive recognition performance. [ABSTRACT FROM AUTHOR]

The millimeter-wave radar can sense the subtle movement of the hand. However, the traditional hand gesture recognition methods do not work well in dynamic interference scenarios. To address this issue, a robust hand gesture recognition method is proposed based on the self-attention time-series neural networks. Firstly, the original radar echo is constructed in terms of frame, sequence, and channel at the input terminal of the network. A one-dimensional time-series neural network is built, and the time-distributed layer is used as the wrapper to extract the feature from each frame sequence independently. Then the self-attention mechanism is employed to assign the adequate weights to the sequence of frames entered in parallel to obtain the inter-frame correlation and suppress the random interference. Finally, the Global AvgPooling layer is used to reduce the number of channels, and the fully connected layer outputs the label of the gesture. The experimental results show that the proposed method can achieve a high recognition rate in the presence of random dynamic interference. [ABSTRACT FROM AUTHOR]

Abstract Radar signal recognition plays a vital role in electronic warfare. For the multifunction radars (MFRs) with complex dynamical modes, the signal recognition needs to identify not only the emitter but also its current functional state. Existing research on MFR recognition mainly focuses on hierarchical modelling approaches. Inspired by recent progress of deep neural networks, the authors propose to further develop radar signal modelling with recurrent neural networks. Here, the authors propose a more efficient method for functional state recognition of MFRs based on a gated recurrent unit (GRU). The method makes full use of the ability of GRUs to automatically learn the characteristics of radar signal sequences, and the nonlinear modelling of GRUs can deal with the corrupted data well. Simulation results on the Mercury radar show that the proposed method achieves a great recognition accuracy with little prior information, and the designed GRU model outperforms the predictive state representation model under the same condition.

Abstract Synthetic wideband signals can obtain a wide spectrum and thus achieve a high range resolution by synthesizing separately emitted signals. A synthetic wideband range profiling method for moving targets based on phase‐derived velocity measurement is proposed. Highly accurate velocity measurement results are obtained by analysing the echo phase information of a moving target, and the accuracy of the velocity estimation results can meet the requirement of the velocity compensation in the synthetic wideband range profiling process. After analysis of the spectrum envelope and the phase of a moving target, a method for synthesizing wideband signals in the frequency domain, thus obtaining a high‐resolution range profile, is developed. The constraint conditions for the motion compensation and synthesis process are analysed and discussed by simulation, and the effectiveness of the proposed method is verified under various circumstances.

For radar target tracking with non‐linear measurements, a sequential linear filtering method is proposed in this study, which includes a linear filter based on the position measurements and a linear sequential filter based on the range rate measurement. The linear radar measurement equation in the sequential filter is estimated based on the position filtering result. Furthermore, the feedback information during tracking is adaptively selected according to the quality of the angle measurements. When tracking the manoeuvring target, the proposed sequential linear filtering with adaptive information feedback is combined with the interacting multiple model (IMM) framework, and the IMM filter based on sequential linear filtering with adaptive model probability selection is proposed. Simulation results demonstrate the effectiveness of the proposed algorithms in non‐manoeuvring and manoeuvring target tracking. Compared with traditional algorithms, the tracking performance is improved by the proposed algorithms in both typical non‐manoeuvring and manoeuvring scenes. [ABSTRACT FROM AUTHOR]

Recent advancements demonstrate the application of machine learning techniques to radar signal processing, particularly the use of neural networks in one-dimensional (1-D) radar data to assist constant false alarm rate (CFAR) based target detection. However, the efficacy of using neural networks alone on 2-D data for target extraction remains to be explored. This research aims to develop a novel approach of designing a convolution neural network (CNN) based target detection system for real radar data. We propose a dual neural network detection scheme in which the first neural net filters out noisy images. Thereafter, we employ a second CNN-based target detector that extracts information from both radar data axes (fast-time and slow-time), thus utilizing additional information for better detection of targets even with low SNR. Moreover, the same CNN, in conjunction with a region-based convolutional network, exploits micro-Doppler features of rotary-wing aircraft in radar returns and uses them to further classify detected targets as fixed-wing or rotary-wing aircraft. To the best of our knowledge, no generalized methodology exists to address these two issues together. The proposed solution replaces the critical CFAR and peak detection algorithms from the radar signal processing chain. This article also introduces a labeled range-Doppler images dataset obtained from real-world air traffic control radar data. Experimental results demonstrate the superior performance of the proposed technique over 1-D and 2-D CFAR and peak detection algorithms. [ABSTRACT FROM AUTHOR]

Copyright of Systems Engineering & Electronics is the property of Journal of Systems Engineering & Electronics Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In the complex electromagnetic environment of cities, detection and positioning of slow and small unmanned aerial vehicles (UAVs) have always been a very challenging research topic. Usually, multiple radar observation stations are required to detect and track targets. Moreover, each radar observation station requires strict synchronisation, and the system is complex and costly. In view of this, under the condition of known prior information of urban geography, this study proposes a method to locate UAVs in complex electromagnetic environments, which use multiple virtual mirror synchronous radars constructed by multipaths and created by the complex buildings in urban environments. First, the simulation model is constructed, and the corresponding simulation parameters are given. Then, the algorithm for solving the single‐station radar location is proposed, and the algorithm is verified by the simulation of a ray propagation model. Finally, the proposed algorithm is further verified based on the measured data captured by the radar system in Section 2. The results show that the proposed algorithm based on the single‐station radar is able to locate the UAV in both line‐of‐sight and non‐line‐of‐sight site, and its positioning accuracy is better than 1 m. [ABSTRACT FROM AUTHOR]

We propose an efficient method to estimate various parameters of a helicopter blade including the number Bnum and the length by separating the blade signal using independent component analysis (ICA) and an analytic solution obtained by monostatic/bistatic geometry. The proposed method is composed of two parts, one for even Bnum and one for odd Bnum. For even Bnum the blade parameters are estimated using ICA plus analytic estimation. For odd Bnum the parameters are estimated using two ICAs plus analytic estimation. Experimental results using the radar signal of a blade model calculated by physical optics demonstrate that the proposed method successfully separates the blade signals for both even and odd Bnums and estimates parameters with high accuracy.

Wave-coefficients (WCs) based target recognition can be theoretically applied to any a frequency range, with increased sensitivity to aspect angle at higher frequencies. When it is utilized to a frequency range in resonance region, the WCs can tolerate fairly big aspect variation and contain enough target information, bringing a competitive advantage in multitudinous target features. Four polarization channel WCs are introduced and independently utilized to identify four aerospace targets under noise and noise-free circumstances. To take full advantage of the WCs in different polarization channels, a novel radar aerospace target recognition method based on multi-polarization channel WCs is proposed in this article. The majority vote rule and a maximum discrepancy rule are combined to identify a target when four polarization channel WCs are available at the same time. Simulation results on four aircraft models show that the proposed technique achieves a recognition performance at least comparable to the best single channel performance.

Global navigation satellite system reflectometry (GNSS-R) has found many applications in the field of Earth observation including ocean wind-speed detection, ice altimetry, soil moisture monitoring, and more. The main focus of GNSS-R research to date has been on forward-scattered reflections, but theoretical work has proposed a backscattering regime and associated new application opportunities, including marine target detection. This article discusses the methods and results of processing the U.K. TechDemoSat-1 raw data collections in a backscattering regime for the first time, with initial results from sea ice datasets presented. The research has also identified a key problem with the backscatter method-for certain geometries the power from the specular point (forward scattered) may contaminate the data. The theory behind this and a method for predicting such occurrences is also discussed.

Convolutional Neural Network (CNN) has been widely applied in the field of synthetic aperture radar (SAR) image recognition. Nevertheless, CNN-based recognition methods usually encounter the problem of poor feature representation ability due to insufficient labeled SAR images. In addition, the large inner-class variety and high cross-class similarity of SAR images pose a challenge for classification. To alleviate the problems mentioned above, we propose a novel few-shot learning (FSL) method for SAR image recognition, which is composed of the multi-feature fusion network (MFFN) and the weighted distance classifier (WDC). The MFFN is utilized to extract input images’ features, and the WDC outputs the classification results based on these features. The MFFN is constructed by adding a multi-scale feature fusion module (MsFFM) and a hand-crafted feature insertion module (HcFIM) to a standard CNN. The feature extraction and representation capability can be enhanced by inserting the traditional hand-crafted features as auxiliary features. With the aid of information from different scales of features, targets of the same class can be more easily aggregated. The weight generation module in WDC is designed to generate category-specific weights for query images. The WDC distributes these weights along the corresponding Euclidean distance to tackle the high cross-class similarity problem. In addition, weight generation loss is proposed to improve recognition performance by guiding the weight generation module. Experimental results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset and the Vehicle and Aircraft (VA) dataset demonstrate that our proposed method surpasses several typical FSL methods.

• To the best of our knowledge, this is the first study to consider the azimuth estimation errors of the HRRP and relate them to deep neural network parameter control. • We proposed two new convolutional kernel parameterization methods, VPCONV and extended CVPCONV, which embed SI into deep network by using it to parameterize the convolutional kernel. This enables direct control of the network structure. Simultaneously, variational encoding is introduced in the reparameterization process to generalize the azimuth estimation error using a heteroscedastic Gaussian distribution with varying mean and variance. Notably, VPCONV and CVPCONV exhibit good transferability and can be used as plug-and-play modules. • Based on CVPCONV, we designed a lightweight network CVPCNN, which can make full use of the azimuth of the HRRP, decouple the tight coupling between the azimuth and target HRRPs, and extract discriminative characteristics that are more relevant to the current environment (azimuth). This helps obtain a more robust and accurate recognition performance. In addition, the CVPCNN introduces a sample-dependent kernel attention mechanism based on parallel convolution to achieve efficient inference, even with increased model complexity. • The effectiveness of the proposed model was tested in a target recognition task using a three-class measured aircraft HRRP dataset. The experiments show that the CVPCNN model, with just three convolutional layers, demonstrate optimal performance on this dataset. This represents a 4 % improvement over the standard CNN model under identical experimental conditions and model structure. Significantly, the CVPCNN maintain good recognition performance under non-ideal conditions, such as small sample and incomplete HRRP attitudes. Radar target recognition tasks often incorporate valuable side information (SI) that aids in the recognition task. The effectiveness of SI has been fully verified in traditional statistical modeling approaches. Its full potential remains untapped in deep learning-based high-resolution range profile (HRRP) target recognition tasks. Taking into account the azimuth sensitivity of HRRP, azimuth is chosen as the SI of the proposed model. To integrate SI into deep neural networks, we propose two novel parameterized convolution methods, namely variational parameterized convolution (VPCONV) and the extended conditionally VPCONV (CVPCONV). Specifically, VPCONV makes the convolutional kernel weights as the function of the auxiliary azimuth, which gives the network the capability of adaptively adjusting to changes in azimuth. Acknowledging the challenge of estimating HRRP azimuth accurately in practical scenarios, VPCONV employs a heteroscedastic Gaussian distribution, featuring varying mean and variance, to generalize the estimation error of SI through variational encoding. Moreover, CVPCONV incorporates the properties of the HRRP samples themselves by kernel attention module. Finally, we present a lightweight SI-based network. The experimental results based on the measured HRRPs validate the effectiveness of proposed method across various extended recognition tasks, underscoring the potential of fusing SI via parameterized convolution in advancing target recognition systems. [ABSTRACT FROM AUTHOR]

When high resolution range profiles(HRRP) are used to recognize radar target, few traditional recognition methods analyse the sparseness of HRRP samples. In order to overcome the tedious analysis problems and simplify the recognition procedure, sparse representation is an effective way to compress HRRP samples and extract the target features. Thus, a federated redundant dictionary and a fast sparse representation algorithm are introduced to implement radar target recognition here. Moreover, a sparse decomposition parameter is adjusted by SNR in order to suppresses noise. The simulation results show that compared with the same kind of RATR algorithms, the algorithm in the paper is practicable, simple and efficient. In contrast to the traditional dimension reduction recognition method, it has better noise robustness and higher recognition ratio.

In order to improve the accuracy and robustness of radar target recognition under low SNR conditions, a novel radar high range resolution profile (HRRP) target recognition method based on feature pyramid fusion lightweight CNN is proposed in this paper. The proposed method combines the multi-scale space theory with a deep convolutional neural network. Because of the local connection characteristic of convolutional kernel, feature extracted by CNN mainly focus on the local information of the target. To make full use of both the local and global information in the target HRRP, multi-scale representation of the HRRP with different Gaussian kernels is introduced to construct the multi-channel input of the model. The generalization performance is improved by reducing the parameters of the proposed model with a depthwise separable convolution feature extraction block. Simultaneously, feature pyramid fusion is adapted to take full advantage of the features extracted by each block, which effectively improves the stability of the model and the training efficiency. The experimental results show that the multi-scale representation of the HRRP contributes to robust feature extraction. Meanwhile, the proposed feature pyramid fusion lightweight CNN can effectively prevent over-fitting and improve the stability of the model.

A novel radar target recognition method based on the deep one-dimensional residual-inception network is proposed for a high-resolution range profile (HRRP). The traditional methods based on shallow models can hardly extract the complete information of the targets HRRP from different angles. The deep models, such as sparse autoencoder, have been adopted to solve this problem. However, these deep models with a huge amount of parameters require more training samples to guarantee the generalization performance. To solve the above-mentioned problem, a model based on the one-dimensional convolutional kernel and a pooling layer is proposed. It is worth noting that the one-dimensional convolutional kernel and pooling operation have the potential to overcome the translation sensitivity and target aspect sensitivity of the HRRP, and both of them can greatly reduce the parameters and improve the generalization performance of the model. In addition, a new loss function is proposed to further enhance the separability of features. The experimental results show that compared with other four deep models, the proposed model can achieve a good performance in recognition accuracy and robustness.

Copyright of Telecommunication Engineering is the property of Telecommunication Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Aiming at the problem of radar small shot target recognition, a comprehensive solution is proposed by combining meta learning and transfer learning, to provide appropriate model learning and classification methods according to different practical application scenarios, so as to improve the efficiency and accuracy of radar small shot target recognition. At the same time, through several groups of comparative experiments, the model performance changes of few shot learning algorithm in the actual radar target recognition scene are deeply analyzed, and two important conclusions that can effectively guide the engineering application are obtained. One is the performance of meta learning model is good when the source task information is sufficient and the difference between the source task and the target task is small, otherwise the transfer learning method is more suitable. The other one is the few shot learning model pay different attention to the external features of radar targets, so the recognition oriented radar imaging should focus on the salient features of the model requirements. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) image classification is one of the most important subjects in automatic target recognition. Therefore, identifying the correct class of targets has significant importance to take a decision. Recently, several deep learning techniques, especially the convolutional neural networks (CNNs), have improved the SAR images classification performance due to its powerful perspective of feature learning and reasoning. Yet, CNN's generally need a huge amount of data for training and do not accurately manage the transformations in the input data. These drawbacks are overcome using a relatively new deep learning approach called capsule networks (CapsNets). In this study, the authors propose a method that adapts and incorporates CapsNet for the SAR image classification problem and improve recognition accuracy through a dual convolution CapsNet framework. Results obtained while experimenting on the moving and stationary target acquisition and recognition data set prove the effectiveness and the robustness of the proposed framework. The proposed experimental results demonstrate the superiority of the employed method overcoming both CNNs and CapsNet separate methods in term of classification accuracy. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) images for automatic target classification (automatic target recognition (ATR)) have attracted significant interest as they can be acquired day and night under a wide range of weather conditions. However, SAR images can be time consuming to analyse, even for experts. ATR can alleviate this burden and deep learning is an attractive solution. A new deep learning Pose‐informed architecture solution, that takes into account the impact of target orientation on the SAR image as the scatterers configuration changes, is proposed. The classification is achieved in two stages. First, the orientation of the target is determined using a Hough transform and a convolutional neural network (CNN). Then, classification is achieved with a CNN specifically trained on targets with similar orientations to the target under test. The networks are trained with translation and SAR‐specific data augmentation. The proposed Pose‐informed deep network architecture was successfully tested on the Military Ground Target Dataset (MGTD) and the Moving and Stationary Target Acquisition and Recognition (MSTAR) datasets. Results show the proposed solution outperformed standard AlexNets on the MGTD, MSTAR extended operating condition (EOC)1, EOC2 and standard operating condition (SOC)10 datasets with a score of 99.13% on the MSTAR SOC10. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) multi‐target interactive motion recognition classifies the type of interactive motion and generates descriptions of the interactive motions at the semantic level by considering the relevance of multi‐target motions. A method for SAR multi‐target interactive motion recognition is proposed, which includes moving target detection, target type recognition, interactive motion feature extraction, and multi‐target interactive motion type recognition. Wavelet thresholding denoising combined with a convolutional neural network (CNN) is proposed for target type recognition. The method performs wavelet thresholding denoising on SAR target images and then uses an eight‐layer CNN named EilNet to achieve target recognition. After target type recognition, a multi‐target interactive motion type recognition method is proposed. A motion feature matrix is constructed for recognition and a four‐layer CNN named FolNet is designed to perform interactive motion type recognition. A motion simulation dataset based on the MSTAR dataset is built, which includes four kinds of interactive motions by two moving targets. The experimental results show that the recognition performance of the authors' Wavelet + EilNet method for target type recognition and FolNet for multi‐target interactive motion type recognition are both better than other methods. Thus, the proposed method is an effective method for SAR multi‐target interactive motion recognition. [ABSTRACT FROM AUTHOR]

We demonstrate and analyze how deploying non-classical intensity correlations obtained from a monolithic semiconductor quantum photon source can enhance classical target detection systems. This is demonstrated by examining the advantages offered by the utilization of the non-classical correlations in a correlation based target detection protocol. We experimentally demonstrate that under the same condition, the target contrast obtained from the protocol when non-classical correlations are utilized exhibits an improvement of up to 17.79 dB over the best classical intensity correlation-based target detection protocol , under 29.69dB channel loss and excess noise 13.40dB stronger than the probe signal. We also assessed how the strong frequency correlations within the non-classical photon pairs can be used to further enhance this protocol. [ABSTRACT FROM AUTHOR]

Motivated by the problem of radar target recognition, we develop a label-aided factor analysis (LA-FA) model for statistical modeling of high-resolution range profile (HRRP) under the prerequisite that the HRRP data are Gaussian distributed. The LA-FA model is the extension of the multitask learning-based factor analysis (MTL-FA) model, which is mainly applied to the recognition problem with small training data size. Compared to the MTL-FA model, our LA-FA model introduces the discrete class labels via Sigmoid-Bernoulli hierarchy to restrict the learning of model parameters, which offers the potential to enhance the separability of statistical models from different classes, thus beneficial to the improvement of recognition capability. In addition, since the noise level of a test sample is usually different from those of the training samples in the real application, we introduce a noise-robust modification method for Gaussian-based models. The proposed modification method is implemented by updating the noise level parameter of the statistical model according to the estimated signal-to-noise ratio (SNR) of test HRRP. Experiments on measured HRRP data demonstrate the better recognition performance of the LA-FA model with limited training data and our noise-robust model modification method under low test SNR. Especially, when there are 20 training HRRP samples per frame, the recognition rate of our LA-FA model is 7% higher than that of the MTL-FA model, and moreover, the recognition accuracy of the noise-robust LA-FA model is 3% higher than that of the LA-FA model without modification under the condition of SNR = 15 dB. [ABSTRACT FROM AUTHOR]

Target recognition of small and slow radar targets with low Radar Target Cross Section (RTCS) on the sea surface is a very important matter in coastal radar field. This article presents a comparative study of the scattered field and the RTCS obtained by analytical means and from computationally simulated small cylindrical radar targets, initially in a homogeneous medium. The forward analysis is based on the simulation of small objects of different sizes, shapes and materials over the sea surface. In this condition, present environment is a heterogeneous medium, the sea - air frontier. The confirmation of the possibility of successful use of numerical simulation in this case is the principal objective of this work. The CST Microwave Studio software programs are used for the simulation; its analysis has required a detailed research over the application of CST in this problem. The obtained results confirm that its solution is the principal task to resolve and to obtain the level of RTCS and field scattering of these objects over sea in each case. Results give us, also, the practical possibility of recognition or identification of each target from the type of material, shapes and sizes of objects. [ABSTRACT FROM AUTHOR]

In this paper, the radar target recognition is given by frequency-diversity RCS (radar cross section) together with kernel scatter difference discrimination. The frequency-diversity technique means to collect electromagnetic signals by sweeping the operation frequencies. Such a technique is usually utilized in inverse scattering and radar target recognition because different frequencies each may contain important information of a target. By using the frequency diversity RCS technique, one can reduce the times of spatial measurement. This is an important contribution since it is always difficult to build a spatial radar measurement in practical battlefield environments. To enhance pattern recognition, the collected RCS data are processed by the kernel scatter difference discrimination, which is improved from the Fisher discrimination. To investigate the capability of tolerating environmental fluctuation, each simulated RCS data is added by a random component prior to implementing pattern recognition. Numerical simulation shows that our recognition scheme is still very accurate even though the RCS contains a random component. [ABSTRACT FROM AUTHOR]

Polarimetric high‐resolution range profile (HRRP) holds great potential for radar automatic target recognition (RATR) owing to its capability of providing both polarimetric and spatial scattering information. In recent years, deep learning (DL) has obtained state‐of‐the‐art results in many classification tasks and has drawn great attention in the RATR field. However, as one of the most challenging tasks in RATR, small training sample case will restrict the application of DL because its superior performance generally depends on a large number of training samples. A feature‐guided deep model based on Transformer framework is proposed for polarimetric HRRP recognition with limited training samples. In the proposed model, artificial features are introduced to the attention module to guide the model focus on the range cells of HRRP with more target scattering information so as to reduce the dependence of the model on the number of training samples. Several different approaches are also studied to measure the similarity between artificial features and HRRP data to further improve the learning capacity of the model. Experimental results demonstrate that the proposed feature‐guided Transformer model modifying by Cosine similarity measure is able to achieve a better performance for polarimetric HRRP recognition with limited training samples. [ABSTRACT FROM AUTHOR]

Target detection is an important part of synthetic aperture radar (SAR) image interpretation, and its performance directly affects the accuracy of localisation and target recognition results in the region of interest. However, the SAR system resolution largely determines the SAR image index and imaging quality, and the target detection performance is closely related to the SAR system resolution in addition to the detection method. To improve the detection performance, the correlation between system resolution and target detection is investigated in this study. Multiple SAR images are used to test a SAR system with reduced resolution obtained by multi-look processing. At the theoretical level, the effect of different resolutions on target detection performance is analysed. MSTAR images, containing image chips with centred targets of varying orientation, are used to analyse the accuracy of the results. By comparing the measured experimental data with the simulation results, it can be obtained that the detection performance is the best when the resolution is roughly equal to the target scale.

Synthetic aperture radar (SAR) target feature extraction is a key technology for SAR target recognition. The existing SAR feature extraction methods mainly focus on amplitude information of the SAR image in spite of phase information. This study proposes a new feature extraction method introducing from concept of wavefront in optics domain. The Zernike polynomials are utilised to approximate the SAR echo wavefront, and these coefficients of Zernike polynomials are used as a new target feature for aiding SAR target recognition. Moreover, an experiment is carried out to prove that this method is effective about a target consisted of two scatters with different spacing between them.