Submit Manuscript
Peer Review
Editor WorkAccepted Article Preview
Abstract
38257KB
Due to the side-looking and coherent imaging mechanisms, feature differences between high-resolution Synthetic Aperture Radar (SAR) images increase when the imaging viewpoint changes considerably, making image registration highly challenging. Traditional registration techniques for high-resolution multiview SAR images mainly face issues, such as insufficient keypoint localization accuracy and low matching precision. This work designs an end-to-end high-resolution multiview SAR image registration network to address the above challenges. The main contributions of this study include the following: A high-resolution SAR image feature extraction method based on a local pixel offset model is proposed. This method introduces a diversity peak loss to guide response weight allocation in the keypoint extraction network and optimizes keypoint coordinates by detecting pixel offsets. A descriptor extraction method is developed based on adaptive adjustment of convolution kernel sampling positions that utilizes sparse cross-entropy loss to supervise descriptor matching in the network. Experimental results show that compared with other registration methods, the proposed algorithm achieves substantial improvements in the high-resolution adjustment of convolution kernel sampling positions, which utilize sparse cross-entropy loss to supervise descriptor matching in the network. Experimental results illustrate that compared with other registration methods, the proposed algorithm achieves remarkable improvements in high-resolution multiview SAR image registration, with an average error reduction of over 65%, 3–5-fold increases in the number of correctly matched point pairs, and an average reduction of over 50% in runtime.
Due to the side-looking and coherent imaging mechanisms, feature differences between high-resolution Synthetic Aperture Radar (SAR) images increase when the imaging viewpoint changes considerably, making image registration highly challenging. Traditional registration techniques for high-resolution multiview SAR images mainly face issues, such as insufficient keypoint localization accuracy and low matching precision. This work designs an end-to-end high-resolution multiview SAR image registration network to address the above challenges. The main contributions of this study include the following: A high-resolution SAR image feature extraction method based on a local pixel offset model is proposed. This method introduces a diversity peak loss to guide response weight allocation in the keypoint extraction network and optimizes keypoint coordinates by detecting pixel offsets. A descriptor extraction method is developed based on adaptive adjustment of convolution kernel sampling positions that utilizes sparse cross-entropy loss to supervise descriptor matching in the network. Experimental results show that compared with other registration methods, the proposed algorithm achieves substantial improvements in the high-resolution adjustment of convolution kernel sampling positions, which utilize sparse cross-entropy loss to supervise descriptor matching in the network. Experimental results illustrate that compared with other registration methods, the proposed algorithm achieves remarkable improvements in high-resolution multiview SAR image registration, with an average error reduction of over 65%, 3–5-fold increases in the number of correctly matched point pairs, and an average reduction of over 50% in runtime.
Aiming to address the problem of increased radar jamming in complex electromagnetic environments and the difficulty of accurately estimating the target signal close to a strong jamming signal, this paper proposes a sparse Direction of Arrival (DOA) estimation method based on Riemann averaging under strong intermittent jamming. First, under the extended coprime array data model, the Riemann averaging is introduced to suppress the jamming signal by leveraging the property that the target signal is continuously active while the strong jamming signal is intermittently active. Then, the covariance matrix of the processed data is vectorized to obtain virtual array reception data. Finally, the sparse iterative covariance-based estimation method, which is used for estimating the DOA under strong intermittent interference, is employed in the virtual domain to reconstruct the sparse signal and estimate the DOA of the target signal. The simulation results show that the method can provide highly accurate DOA estimation for weak target signals whose angles are closely adjacent to strong interference signals when the number of signal sources is unknown. Compared with existing subspace algorithms and sparse reconstruction class algorithms, the proposed algorithm has higher estimation accuracy and angular resolution at a smaller number of snapshots, as well as a lower signal-to-noise ratio.
Aiming to address the problem of increased radar jamming in complex electromagnetic environments and the difficulty of accurately estimating the target signal close to a strong jamming signal, this paper proposes a sparse Direction of Arrival (DOA) estimation method based on Riemann averaging under strong intermittent jamming. First, under the extended coprime array data model, the Riemann averaging is introduced to suppress the jamming signal by leveraging the property that the target signal is continuously active while the strong jamming signal is intermittently active. Then, the covariance matrix of the processed data is vectorized to obtain virtual array reception data. Finally, the sparse iterative covariance-based estimation method, which is used for estimating the DOA under strong intermittent interference, is employed in the virtual domain to reconstruct the sparse signal and estimate the DOA of the target signal. The simulation results show that the method can provide highly accurate DOA estimation for weak target signals whose angles are closely adjacent to strong interference signals when the number of signal sources is unknown. Compared with existing subspace algorithms and sparse reconstruction class algorithms, the proposed algorithm has higher estimation accuracy and angular resolution at a smaller number of snapshots, as well as a lower signal-to-noise ratio.
Land-sea clutter classification is essential for boosting the target positioning accuracy of skywave over-the-horizon radar. This classification process involves discriminating whether each azimuth-range cell in the Range-Doppler (RD) map is overland or sea. Traditional deep learning methods for this task require extensive, high-quality, and class-balanced labeled samples, leading to long training periods and high costs. In addition, these methods typically use a single azimuth-range cell clutter without considering intra-class and inter-class relationships, resulting in poor model performance. To address these challenges, this study analyzes the correlation between adjacent azimuth-range cells, and converts land-sea clutter data from Euclidean space into graph data in non-Euclidean space, thereby incorporating sample relationships. We propose a Multi-Channel Graph Convolutional Networks (MC-GCN) for land-sea clutter classification. MC-GCN decomposes graph data from a single channel into multiple channels, each containing a single type of edge and a weight matrix. This approach restricts node information aggregation, effectively reducing node attribute misjudgment caused by data heterogeneity. For validation, RD maps from various seasons, times, and detection areas were selected. Based on radar parameters, data characteristics, and sample proportions, we construct a land-sea clutter original dataset containing 12 different scenes and a land-sea clutter scarce dataset containing 36 different configurations. The effectiveness of MC-GCN is confirmed, with the approach outperforming state-of-the-art classification methods with a classification accuracy of at least 92%.
Land-sea clutter classification is essential for boosting the target positioning accuracy of skywave over-the-horizon radar. This classification process involves discriminating whether each azimuth-range cell in the Range-Doppler (RD) map is overland or sea. Traditional deep learning methods for this task require extensive, high-quality, and class-balanced labeled samples, leading to long training periods and high costs. In addition, these methods typically use a single azimuth-range cell clutter without considering intra-class and inter-class relationships, resulting in poor model performance. To address these challenges, this study analyzes the correlation between adjacent azimuth-range cells, and converts land-sea clutter data from Euclidean space into graph data in non-Euclidean space, thereby incorporating sample relationships. We propose a Multi-Channel Graph Convolutional Networks (MC-GCN) for land-sea clutter classification. MC-GCN decomposes graph data from a single channel into multiple channels, each containing a single type of edge and a weight matrix. This approach restricts node information aggregation, effectively reducing node attribute misjudgment caused by data heterogeneity. For validation, RD maps from various seasons, times, and detection areas were selected. Based on radar parameters, data characteristics, and sample proportions, we construct a land-sea clutter original dataset containing 12 different scenes and a land-sea clutter scarce dataset containing 36 different configurations. The effectiveness of MC-GCN is confirmed, with the approach outperforming state-of-the-art classification methods with a classification accuracy of at least 92%.
In non-inductive radar vital sign monitoring, frequency-modulated radars (such as Frequency Modulated Continuous Wave (FMCW) and Ultra WideBand (UWB)) are more effective than Continuous Wave (CW) radars at distinguishing targets from clutter in terms of distance. Using range Fourier transform, the heartbeat and breathing signals can be extracted from quasi-static targets across various distance intervals, thereby improving monitoring accuracy. However, the commonly used range Fast Fourier Transform (FFT) presents certain limitations: The breathing amplitude of the subject may cross the range bin boundary, compromising signal integrity, while breathing movements can cause amplitude modulation of physiological signals, hindering waveform recovery. To address these reasons, we propose an algorithm architecture featuring range tap reconstruction and dynamic demodulation. We tested the algorithm performance in simulations and experiments for the cross range bin cases. Simulation results indicate that processing signals crossing range bins with our algorithm improves the signal-to-noise ratio by 17±5 dB. In addition, experiments recorded Doppler Heartbeat Diagram (DHD) signals from eight subjects, comparing the consistency between the DHD signals and the ballistocardiogram. The root means square error of the C-C interval in the DHD signal relative to the J-J interval in the BallistoCardioGram (BCG) signal was 21.58±13.26 ms (3.40%±2.08%).
In non-inductive radar vital sign monitoring, frequency-modulated radars (such as Frequency Modulated Continuous Wave (FMCW) and Ultra WideBand (UWB)) are more effective than Continuous Wave (CW) radars at distinguishing targets from clutter in terms of distance. Using range Fourier transform, the heartbeat and breathing signals can be extracted from quasi-static targets across various distance intervals, thereby improving monitoring accuracy. However, the commonly used range Fast Fourier Transform (FFT) presents certain limitations: The breathing amplitude of the subject may cross the range bin boundary, compromising signal integrity, while breathing movements can cause amplitude modulation of physiological signals, hindering waveform recovery. To address these reasons, we propose an algorithm architecture featuring range tap reconstruction and dynamic demodulation. We tested the algorithm performance in simulations and experiments for the cross range bin cases. Simulation results indicate that processing signals crossing range bins with our algorithm improves the signal-to-noise ratio by 17±5 dB. In addition, experiments recorded Doppler Heartbeat Diagram (DHD) signals from eight subjects, comparing the consistency between the DHD signals and the ballistocardiogram. The root means square error of the C-C interval in the DHD signal relative to the J-J interval in the BallistoCardioGram (BCG) signal was 21.58±13.26 ms (3.40%±2.08%).
Imaging of passive jamming objects has been a hot topic in radar imaging and countermeasures research, which directly affects the detection and recognition capabilities of radar targets. In the microwave band, the long dwell time required to generate a single image with desired azimuthal resolution makes it difficult to directly distinguish passive jamming objects based on imaging results. In addition, there is a lack of time-dimensional resolution. In comparison, terahertz imaging systems require a shorter synthetic aperture to achieve the same azimuthal resolution, making it easier to obtain low-latency, high-resolution, and high-frame-rate imaging results. Hence, terahertz radar has considerable potential in Video Synthetic Aperture Radar (ViSAR) technology. First, the aperture division and imaging resolutions of airborne terahertz ViSAR are briefly analyzed. Subsequently, imaging results and characteristics of stationary passive jamming objects, such as corner reflector arrays and camouflage mats, are explored before and after motion compensation. Further, the phenomenon that camouflage mats with fluctuating grids exhibit roughness in the terahertz band is demonstrated, exhibiting the special scattering characteristics of the terahertz band. Next, considering rotating corner reflectors as an example of moving passive jamming objects, their characteristics regarding suppressive interference are analyzed. Considering that stationary scenes feature similarity under adjacent apertures, rotating corner reflectors can be directly detected by incoherent image subtraction after inter-frame image and amplitude registrations, followed by the extraction of signals of interest and non-parametrical compensation. Currently, few field experiments regarding the imaging of passive jamming objects using terahertz ViSAR are being reported. Airborne field experiments have been performed to effectively demonstrate the high-resolution and high-frame-rate imaging capabilities of terahertz ViSAR
Imaging of passive jamming objects has been a hot topic in radar imaging and countermeasures research, which directly affects the detection and recognition capabilities of radar targets. In the microwave band, the long dwell time required to generate a single image with desired azimuthal resolution makes it difficult to directly distinguish passive jamming objects based on imaging results. In addition, there is a lack of time-dimensional resolution. In comparison, terahertz imaging systems require a shorter synthetic aperture to achieve the same azimuthal resolution, making it easier to obtain low-latency, high-resolution, and high-frame-rate imaging results. Hence, terahertz radar has considerable potential in Video Synthetic Aperture Radar (ViSAR) technology. First, the aperture division and imaging resolutions of airborne terahertz ViSAR are briefly analyzed. Subsequently, imaging results and characteristics of stationary passive jamming objects, such as corner reflector arrays and camouflage mats, are explored before and after motion compensation. Further, the phenomenon that camouflage mats with fluctuating grids exhibit roughness in the terahertz band is demonstrated, exhibiting the special scattering characteristics of the terahertz band. Next, considering rotating corner reflectors as an example of moving passive jamming objects, their characteristics regarding suppressive interference are analyzed. Considering that stationary scenes feature similarity under adjacent apertures, rotating corner reflectors can be directly detected by incoherent image subtraction after inter-frame image and amplitude registrations, followed by the extraction of signals of interest and non-parametrical compensation. Currently, few field experiments regarding the imaging of passive jamming objects using terahertz ViSAR are being reported. Airborne field experiments have been performed to effectively demonstrate the high-resolution and high-frame-rate imaging capabilities of terahertz ViSAR
The miniature multistatic Synthetic Aperture Radar (SAR) system uses a flexible configuration of transceiver division compared with the miniature monostatic SAR system, thereby affording the advantages of multi-angle imaging. As the transceiver-separated SAR system uses mutually independent oscillator sources, phase synchronization is necessary for high-precision imaging of the miniature multistatic SAR. Although current research on phase synchronization schemes for bistatic SAR is relatively mature, these schemes are primarily based on the pulse SAR system. However, a paucity of research exists on phase synchronization for the miniature multistatic Frequency Modulated Continuous Wave (FMCW) SAR. In comparison with the pulse SAR, the FMCW SAR system lacks a temporal interval between the transmitted pulses. Consequently, some phase synchronization schemes developed for the pulse SAR system cannot be directly applied to the FMCW SAR system. To this end, this study proposes a novel phase synchronization method for the miniature multistatic FMCW SAR, effectively resolving the problem of the FMCW SAR. This method uses the generalized Short-Time Shift-Orthogonal (STSO) waveform as the phase synchronization signal of disparate radar platforms. The phase error between the radar platforms can be effectively extracted through pulse compression to realize phase synchronization. Compared with the conventional linear frequency-modulated waveform, after the generalized STSO waveform is pulsed by the same pulse compression function, the interference signal energy is concentrated away from the peak of the matching signal and the phase synchronization accuracy is enhanced. Furthermore, the proposed method is adapted to the characteristics of dechirp reception in FMCW miniature multistatic SAR systems, and ground and numerical simulation experiments verify that the proposed method has high synchronization accuracy.
The miniature multistatic Synthetic Aperture Radar (SAR) system uses a flexible configuration of transceiver division compared with the miniature monostatic SAR system, thereby affording the advantages of multi-angle imaging. As the transceiver-separated SAR system uses mutually independent oscillator sources, phase synchronization is necessary for high-precision imaging of the miniature multistatic SAR. Although current research on phase synchronization schemes for bistatic SAR is relatively mature, these schemes are primarily based on the pulse SAR system. However, a paucity of research exists on phase synchronization for the miniature multistatic Frequency Modulated Continuous Wave (FMCW) SAR. In comparison with the pulse SAR, the FMCW SAR system lacks a temporal interval between the transmitted pulses. Consequently, some phase synchronization schemes developed for the pulse SAR system cannot be directly applied to the FMCW SAR system. To this end, this study proposes a novel phase synchronization method for the miniature multistatic FMCW SAR, effectively resolving the problem of the FMCW SAR. This method uses the generalized Short-Time Shift-Orthogonal (STSO) waveform as the phase synchronization signal of disparate radar platforms. The phase error between the radar platforms can be effectively extracted through pulse compression to realize phase synchronization. Compared with the conventional linear frequency-modulated waveform, after the generalized STSO waveform is pulsed by the same pulse compression function, the interference signal energy is concentrated away from the peak of the matching signal and the phase synchronization accuracy is enhanced. Furthermore, the proposed method is adapted to the characteristics of dechirp reception in FMCW miniature multistatic SAR systems, and ground and numerical simulation experiments verify that the proposed method has high synchronization accuracy.
Human pose estimation holds tremendous potential in fields such as human-computer interaction, motion capture, and virtual reality, making it a focus in human perception research. However, optical image-based pose estimation methods are often limited by lighting conditions and privacy concerns. Therefore, the use of wireless signals that can operate under various lighting conditions and obstructions while ensuring privacy is gaining increasing attention for human pose estimation. Wireless signal-based pose estimation technologies can be categorized into high-frequency and low-frequency methods. These methods differ in their hardware systems, signal characteristics, noise processing, and deep learning algorithm design based on the signal frequency used. This paper highlights research advancements and notable works in human pose reconstruction using millimeter-wave radar, through-wall radar, and WiFi. It analyzes the advantages and limitations of each signal type and explores potential research challenges and future developments in the field.
Human pose estimation holds tremendous potential in fields such as human-computer interaction, motion capture, and virtual reality, making it a focus in human perception research. However, optical image-based pose estimation methods are often limited by lighting conditions and privacy concerns. Therefore, the use of wireless signals that can operate under various lighting conditions and obstructions while ensuring privacy is gaining increasing attention for human pose estimation. Wireless signal-based pose estimation technologies can be categorized into high-frequency and low-frequency methods. These methods differ in their hardware systems, signal characteristics, noise processing, and deep learning algorithm design based on the signal frequency used. This paper highlights research advancements and notable works in human pose reconstruction using millimeter-wave radar, through-wall radar, and WiFi. It analyzes the advantages and limitations of each signal type and explores potential research challenges and future developments in the field.
In recent years, target recognition systems based on radar sensor networks have been widely studied in the field of automatic target recognition. These systems observe the target from multiple angles to achieve robust recognition, which also brings the problem of using the correlation and difference information of multiradar sensor echo data. Furthermore, most existing studies used large-scale labeled data to obtain prior knowledge of the target. Considering that a large amount of unlabeled data is not effectively used in target recognition tasks, this paper proposes an HRRP unsupervised target feature extraction method based on Multiple Contrastive Loss (MCL) in radar sensor networks. The proposed method combines instance level loss, Fisher loss, and semantic consistency loss constraints to identify consistent and discriminative feature vectors among the echoes of multiple radar sensors and then use them in subsequent target recognition tasks. Specifically, the original echo data are mapped to the contrast loss space and the semantic label space. In the contrast loss space, the contrastive loss is used to constrain the similarity and aggregation of samples so that the relative and absolute distances between different echoes of the same target obtained by different sensors are reduced while the relative and absolute distances between different target echoes are increased. In the semantic loss space, the extracted discriminant features are used to constrain the semantic labels so that the semantic information and discriminant features are consistent. Experiments on an actual civil aircraft dataset revealed that the target recognition accuracy of the MCL-based method is improved by 0.4% and 1.4%, respectively, compared with the most advanced unsupervised algorithm CC and supervised target recognition algorithm PNN. Further, MCL can effectively improve the target recognition performance of radar sensors when applied in conjunction with the sensors.
In recent years, target recognition systems based on radar sensor networks have been widely studied in the field of automatic target recognition. These systems observe the target from multiple angles to achieve robust recognition, which also brings the problem of using the correlation and difference information of multiradar sensor echo data. Furthermore, most existing studies used large-scale labeled data to obtain prior knowledge of the target. Considering that a large amount of unlabeled data is not effectively used in target recognition tasks, this paper proposes an HRRP unsupervised target feature extraction method based on Multiple Contrastive Loss (MCL) in radar sensor networks. The proposed method combines instance level loss, Fisher loss, and semantic consistency loss constraints to identify consistent and discriminative feature vectors among the echoes of multiple radar sensors and then use them in subsequent target recognition tasks. Specifically, the original echo data are mapped to the contrast loss space and the semantic label space. In the contrast loss space, the contrastive loss is used to constrain the similarity and aggregation of samples so that the relative and absolute distances between different echoes of the same target obtained by different sensors are reduced while the relative and absolute distances between different target echoes are increased. In the semantic loss space, the extracted discriminant features are used to constrain the semantic labels so that the semantic information and discriminant features are consistent. Experiments on an actual civil aircraft dataset revealed that the target recognition accuracy of the MCL-based method is improved by 0.4% and 1.4%, respectively, compared with the most advanced unsupervised algorithm CC and supervised target recognition algorithm PNN. Further, MCL can effectively improve the target recognition performance of radar sensors when applied in conjunction with the sensors.
The ionosphere can distort received signals, degrade imaging quality, and decrease interferometric and polarimetric accuracies of spaceborne Synthetic Aperture Radars (SAR). The low-frequency systems operating at L-band and P-band are very susceptible to such problems. From another viewpoint, low-frequency spaceborne SARs can capture ionospheric structures with different spatial scales over the observed scope, and their echo and image data have sufficient ionospheric information, offering great probability for high-precision and high-resolution ionospheric probing. The research progress of ionospheric probing based on spaceborne SARs is reviewed in this paper. The technological system of this field is summarized from three aspects: Mapping of background ionospheric total electron content, tomography of ionospheric electron density, and probing of ionospheric irregularities. The potential of the low-frequency spaceborne SARs in mapping ionospheric local refined structures and global tendency is emphasized, and the future development direction is prospected.
The ionosphere can distort received signals, degrade imaging quality, and decrease interferometric and polarimetric accuracies of spaceborne Synthetic Aperture Radars (SAR). The low-frequency systems operating at L-band and P-band are very susceptible to such problems. From another viewpoint, low-frequency spaceborne SARs can capture ionospheric structures with different spatial scales over the observed scope, and their echo and image data have sufficient ionospheric information, offering great probability for high-precision and high-resolution ionospheric probing. The research progress of ionospheric probing based on spaceborne SARs is reviewed in this paper. The technological system of this field is summarized from three aspects: Mapping of background ionospheric total electron content, tomography of ionospheric electron density, and probing of ionospheric irregularities. The potential of the low-frequency spaceborne SARs in mapping ionospheric local refined structures and global tendency is emphasized, and the future development direction is prospected.
Passive radar plays an important role in early warning detection and Low Slow Small (LSS) target detection. Due to the uncontrollable source of passive radar signal radiations, target characteristics are more complex, which makes target detection and identification extremely difficult. In this paper, a passive radar LSS detection dataset (LSS-PR-1.0) is constructed, which contains the radar echo signals of four typical sea and air targets, namely helicopters, unmanned aerial vehicles, speedboats, and passenger ships, as well as sea clutter data at low and high sea states. It provides data support for radar research. In terms of target feature extraction and analysis, the singular-value-decomposition sea-clutter-suppression method is first adopted to remove the influence of the strong Bragg peak of sea clutter on target echo. On this basis, four categories of ten multi-domain feature extraction and analysis methods are proposed, including time-domain features (relative average amplitude), frequency-domain features (spectral features, Doppler waterfall plot, and range Doppler features), time-frequency-domain features, and motion features (heading difference, trajectory parameters, speed variation interval, speed variation coefficient, and acceleration). Based on the actual measurement data, a comparative analysis is conducted on the characteristics of four types of sea and air targets, summarizing the patterns of various target characteristics and laying the foundation for subsequent target recognition.
Passive radar plays an important role in early warning detection and Low Slow Small (LSS) target detection. Due to the uncontrollable source of passive radar signal radiations, target characteristics are more complex, which makes target detection and identification extremely difficult. In this paper, a passive radar LSS detection dataset (LSS-PR-1.0) is constructed, which contains the radar echo signals of four typical sea and air targets, namely helicopters, unmanned aerial vehicles, speedboats, and passenger ships, as well as sea clutter data at low and high sea states. It provides data support for radar research. In terms of target feature extraction and analysis, the singular-value-decomposition sea-clutter-suppression method is first adopted to remove the influence of the strong Bragg peak of sea clutter on target echo. On this basis, four categories of ten multi-domain feature extraction and analysis methods are proposed, including time-domain features (relative average amplitude), frequency-domain features (spectral features, Doppler waterfall plot, and range Doppler features), time-frequency-domain features, and motion features (heading difference, trajectory parameters, speed variation interval, speed variation coefficient, and acceleration). Based on the actual measurement data, a comparative analysis is conducted on the characteristics of four types of sea and air targets, summarizing the patterns of various target characteristics and laying the foundation for subsequent target recognition.
Amidst the global aging trend and a growing emphasis on healthy living, there is an increased demand for unobtrusive home health monitoring systems. However, the current mainstream detection methods in this regard suffer from low privacy trust, poor electromagnetic compatibility, and high manufacturing costs. To address these challenges, this paper introduces a noncontact vital signal collection device using Ultrasonic radar (U-Sodar), including a set of hardware based on a three-transmitter four-receiver Multiple Input Multiple Output (MIMO) architecture and a set of signal processing algorithms. The U-Sodar local oscillator uses frequency division technology with low phase noise and high detection accuracy; the receiver employs front-end direct sampling technology to simplify the involved structure and effectively reduce external noise, and the transmitter uses an adjustable PWM direct drive to emit various ultrasonic waveforms, possessing software-defined ultrasonic system characteristics. The signal processing algorithm of U-Sodar adopts the graph processing technique of signal chord length and realizes accurate recovery of signal phase under 5 dB Signal-to-noise ratio (SNR) using picture filtering and then reconstruction. Experimental tests on the U-Sodar system demonstrated its anti-interference and penetration capabilities, proving that ultrasonic penetration relies on material porosity rather than intermedium vibration conduction. The minimum measurable displacement for a given SNR with correct demodulation probability is also derived. The results of actual human vital sign signal measurement experiments indicate that U-Sodar can accurately measure respiration and heartbeat at 3.0 m and 1.5 m, respectively, and the heartbeat waveforms can be measured within 1.0 m. Overall, the experimental results demonstrate the feasibility and application potential of U-Sodar in noncontact vital sign detection.
Amidst the global aging trend and a growing emphasis on healthy living, there is an increased demand for unobtrusive home health monitoring systems. However, the current mainstream detection methods in this regard suffer from low privacy trust, poor electromagnetic compatibility, and high manufacturing costs. To address these challenges, this paper introduces a noncontact vital signal collection device using Ultrasonic radar (U-Sodar), including a set of hardware based on a three-transmitter four-receiver Multiple Input Multiple Output (MIMO) architecture and a set of signal processing algorithms. The U-Sodar local oscillator uses frequency division technology with low phase noise and high detection accuracy; the receiver employs front-end direct sampling technology to simplify the involved structure and effectively reduce external noise, and the transmitter uses an adjustable PWM direct drive to emit various ultrasonic waveforms, possessing software-defined ultrasonic system characteristics. The signal processing algorithm of U-Sodar adopts the graph processing technique of signal chord length and realizes accurate recovery of signal phase under 5 dB Signal-to-noise ratio (SNR) using picture filtering and then reconstruction. Experimental tests on the U-Sodar system demonstrated its anti-interference and penetration capabilities, proving that ultrasonic penetration relies on material porosity rather than intermedium vibration conduction. The minimum measurable displacement for a given SNR with correct demodulation probability is also derived. The results of actual human vital sign signal measurement experiments indicate that U-Sodar can accurately measure respiration and heartbeat at 3.0 m and 1.5 m, respectively, and the heartbeat waveforms can be measured within 1.0 m. Overall, the experimental results demonstrate the feasibility and application potential of U-Sodar in noncontact vital sign detection.
Unmanned Aerial Vehicle (UAV)-borne radar technology can solve the problems associated with noncontact vital sign sensing, such as limited detection range, slow moving speed, and difficult access to certain areas. In this study, we mount a 4D imaging radar on a multirotor UAV and propose a UAV-borne radar-based method for sensing vital signs through point cloud registration. Through registration and motion compensation of the radar point cloud, the motion error interference of UAV hovering is eliminated; vital sign signals are then obtained after aligning the human target. Simulation results show that the proposed method can effectively align the 4D radar point cloud sequence and accurately extract the respiration and heartbeat signals of human targets, thereby providing a way to realize UAV-borne vital sign sensing.
Unmanned Aerial Vehicle (UAV)-borne radar technology can solve the problems associated with noncontact vital sign sensing, such as limited detection range, slow moving speed, and difficult access to certain areas. In this study, we mount a 4D imaging radar on a multirotor UAV and propose a UAV-borne radar-based method for sensing vital signs through point cloud registration. Through registration and motion compensation of the radar point cloud, the motion error interference of UAV hovering is eliminated; vital sign signals are then obtained after aligning the human target. Simulation results show that the proposed method can effectively align the 4D radar point cloud sequence and accurately extract the respiration and heartbeat signals of human targets, thereby providing a way to realize UAV-borne vital sign sensing.
Due to their many advantages, such as simple structure, low transmission power, strong penetration capability, high resolution, and high transmission speed, UWB (Ultra-Wide Band) radars have been widely used for detecting life information in various scenarios. To effectively detect life information, the key is to use radar echo information–processing technology to extract the breathing and heartbeat signals of the involved person from UWB radar echoes. This technology is crucial for determining life information in different scenarios, such as obtaining location information, monitoring and preventing diseases, and ensuring personnel safety. Therefore, this paper introduces a UWB radar and its classification, electromagnetic scattering mechanisms, and detection principles. It also analyzes the current state of radar echo model construction for breathing and heartbeat signals. The paper then reviews existing methods for extracting breathing and heartbeat signals, including time domain, frequency domain, and time–frequency domain analysis methods. Finally, it summarizes research progress in breathing and heartbeat signal extraction in various scenarios, such as mine rescue, earthquake rescue, medical health, and through-wall detection, as well as the main problems in current research and focus areas for future research.
Due to their many advantages, such as simple structure, low transmission power, strong penetration capability, high resolution, and high transmission speed, UWB (Ultra-Wide Band) radars have been widely used for detecting life information in various scenarios. To effectively detect life information, the key is to use radar echo information–processing technology to extract the breathing and heartbeat signals of the involved person from UWB radar echoes. This technology is crucial for determining life information in different scenarios, such as obtaining location information, monitoring and preventing diseases, and ensuring personnel safety. Therefore, this paper introduces a UWB radar and its classification, electromagnetic scattering mechanisms, and detection principles. It also analyzes the current state of radar echo model construction for breathing and heartbeat signals. The paper then reviews existing methods for extracting breathing and heartbeat signals, including time domain, frequency domain, and time–frequency domain analysis methods. Finally, it summarizes research progress in breathing and heartbeat signal extraction in various scenarios, such as mine rescue, earthquake rescue, medical health, and through-wall detection, as well as the main problems in current research and focus areas for future research.
Ultra-WideBand (UWB) radar exhibits strong antijamming capabilities and high penetrability, making it widely used for through-wall human-target detection. Although single-transmitter, single-receiver radar offers the advantages of a compact size and lightweight design, it cannot achieve Two-Dimensional (2D) target localization. Multiple-Input Multiple-Output (MIMO) array radar can localize targets but faces a trade-off between size and resolution and involves longer computation durations. This paper proposes an automatic multitarget detection method based on distributed through-wall radar. First, the echo signal is preprocessed in the time domain and then transformed into the time-frequency domain. Target candidate distance cells are identified using a constant false alarm rate detection method, and candidate signals are enhanced using a filtering matrix. The enhanced signals are then correlated based on vital information, such as breathing, to achieve target matching. Finally, a positioning module is employed to determine the radar’s location, enabling rapid and automatic detection of the target’s location. To mitigate the effect of occasional errors on the final positioning results, a scene segmentation method is used to achieve 2D localization of human targets in through-wall scenarios. Experimental results demonstrate that the proposed method can successfully detect and localize multiple targets in through-wall scenarios, with a computation duration of 0.95 s based on the measured data. In particular, the method is over four times faster than other methods.
Ultra-WideBand (UWB) radar exhibits strong antijamming capabilities and high penetrability, making it widely used for through-wall human-target detection. Although single-transmitter, single-receiver radar offers the advantages of a compact size and lightweight design, it cannot achieve Two-Dimensional (2D) target localization. Multiple-Input Multiple-Output (MIMO) array radar can localize targets but faces a trade-off between size and resolution and involves longer computation durations. This paper proposes an automatic multitarget detection method based on distributed through-wall radar. First, the echo signal is preprocessed in the time domain and then transformed into the time-frequency domain. Target candidate distance cells are identified using a constant false alarm rate detection method, and candidate signals are enhanced using a filtering matrix. The enhanced signals are then correlated based on vital information, such as breathing, to achieve target matching. Finally, a positioning module is employed to determine the radar’s location, enabling rapid and automatic detection of the target’s location. To mitigate the effect of occasional errors on the final positioning results, a scene segmentation method is used to achieve 2D localization of human targets in through-wall scenarios. Experimental results demonstrate that the proposed method can successfully detect and localize multiple targets in through-wall scenarios, with a computation duration of 0.95 s based on the measured data. In particular, the method is over four times faster than other methods.
Since 2010, the utilization of commercial WiFi devices for contact-free respiration monitoring has garnered significant attention. However, existing WiFi-based respiration detection methods are susceptible to constraints imposed by hardware limitations and require the person to directly face the WiFi device. Specifically, signal reflection from the thoracic cavity diminishes when the body is oriented sideways or with the back toward the device, leading to complexities in respiratory monitoring. To mitigate these hardware-associated limitations and enhance robustness, we leveraged the signal-amplifying potential of Intelligent Reflecting Surfaces (IRS) to establish a high-precision respiration detection system. This system capitalizes on IRS technology to manipulate signal propagation within the environment to enhance signal reflection from the body, finally achieving posture-resilient respiratory monitoring. Furthermore, the system can be easily deployed without the prior knowledge of antenna placement or environmental intricacies. Compared with conventional techniques, our experimental results validate that this system markedly enhances respiratory monitoring across various postural configurations in indoor environments.
Since 2010, the utilization of commercial WiFi devices for contact-free respiration monitoring has garnered significant attention. However, existing WiFi-based respiration detection methods are susceptible to constraints imposed by hardware limitations and require the person to directly face the WiFi device. Specifically, signal reflection from the thoracic cavity diminishes when the body is oriented sideways or with the back toward the device, leading to complexities in respiratory monitoring. To mitigate these hardware-associated limitations and enhance robustness, we leveraged the signal-amplifying potential of Intelligent Reflecting Surfaces (IRS) to establish a high-precision respiration detection system. This system capitalizes on IRS technology to manipulate signal propagation within the environment to enhance signal reflection from the body, finally achieving posture-resilient respiratory monitoring. Furthermore, the system can be easily deployed without the prior knowledge of antenna placement or environmental intricacies. Compared with conventional techniques, our experimental results validate that this system markedly enhances respiratory monitoring across various postural configurations in indoor environments.
Sleep Apnea Hypopnea Syndrome (SAHS) is a common chronic sleep-related breathing disorder that affects individuals’ sleep quality and physical health. This article presents a sleep apnea and hypopnea detection framework based on multisource signal fusion. Integrating millimeter-wave radar micro-motion signals and pulse wave signals of PhotoPlethysmoGraphy (PPG) achieves a highly reliable and light-contact diagnosis of SAHS, addressing the drawbacks of traditional medical methods that rely on PolySomnoGraphy (PSG) for sleep monitoring, such as poor comfort and high costs. This study used a radar and pulse wave data preprocessing algorithm to extract time-frequency information and artificial features from the signals, balancing the accuracy and robustness of sleep-breathing abnormality event detection Additionally, a deep neural network was designed to fuse the two types of signals for precise identification of sleep apnea and hypopnea events, and to estimate the Apnea-Hypopnea Index (AHI) for quantitative assessment of sleep-breathing abnormality severity. Experimental results of a clinical trial dataset at Shanghai Jiaotong University School of Medicine Affiliated Sixth People’s Hospital demonstrated that the AHI estimated by the proposed approach correlates with the gold standard PSG with a coefficient of 0.93, indicating good consistency. This approach is a promiseing tool for home sleep-breathing monitoring and preliminary diagnosis of SAHS.
Sleep Apnea Hypopnea Syndrome (SAHS) is a common chronic sleep-related breathing disorder that affects individuals’ sleep quality and physical health. This article presents a sleep apnea and hypopnea detection framework based on multisource signal fusion. Integrating millimeter-wave radar micro-motion signals and pulse wave signals of PhotoPlethysmoGraphy (PPG) achieves a highly reliable and light-contact diagnosis of SAHS, addressing the drawbacks of traditional medical methods that rely on PolySomnoGraphy (PSG) for sleep monitoring, such as poor comfort and high costs. This study used a radar and pulse wave data preprocessing algorithm to extract time-frequency information and artificial features from the signals, balancing the accuracy and robustness of sleep-breathing abnormality event detection Additionally, a deep neural network was designed to fuse the two types of signals for precise identification of sleep apnea and hypopnea events, and to estimate the Apnea-Hypopnea Index (AHI) for quantitative assessment of sleep-breathing abnormality severity. Experimental results of a clinical trial dataset at Shanghai Jiaotong University School of Medicine Affiliated Sixth People’s Hospital demonstrated that the AHI estimated by the proposed approach correlates with the gold standard PSG with a coefficient of 0.93, indicating good consistency. This approach is a promiseing tool for home sleep-breathing monitoring and preliminary diagnosis of SAHS.
In recent years, there has been an increasing interest in respiratory monitoring in multiperson environments and simultaneous monitoring of the health status of multiple people. Among the algorithms developed for multiperson respiratory detection, blind source separation algorithms have attracted the attention of researchers because they do not require prior information and are less dependent on hardware performance. However, in the context of multiperson respiratory monitoring, the current blind source separation algorithm usually separates phase signals as the source signal. This article compares the distance dimension and phase signals under Frequency-modulated continuous-wave radar, calculates the approximate error associated with using the phase signal as the source signal, and verifies the separation effect through simulations. The distance dimension signal is better to use as the source signal. In addition, this article proposes a multiperson respiratory signal separation algorithm based on noncircular complex independent component analysis and analyzes the impact of different respiratory signal parameters on the separation effect. Simulation and experimental measurements show that the proposed method is suitable for detecting multiperson respiratory signals under controlled conditions and can accurately separate respiratory signals when the angle of the two targets to the radar is 9.46°.
In recent years, there has been an increasing interest in respiratory monitoring in multiperson environments and simultaneous monitoring of the health status of multiple people. Among the algorithms developed for multiperson respiratory detection, blind source separation algorithms have attracted the attention of researchers because they do not require prior information and are less dependent on hardware performance. However, in the context of multiperson respiratory monitoring, the current blind source separation algorithm usually separates phase signals as the source signal. This article compares the distance dimension and phase signals under Frequency-modulated continuous-wave radar, calculates the approximate error associated with using the phase signal as the source signal, and verifies the separation effect through simulations. The distance dimension signal is better to use as the source signal. In addition, this article proposes a multiperson respiratory signal separation algorithm based on noncircular complex independent component analysis and analyzes the impact of different respiratory signal parameters on the separation effect. Simulation and experimental measurements show that the proposed method is suitable for detecting multiperson respiratory signals under controlled conditions and can accurately separate respiratory signals when the angle of the two targets to the radar is 9.46°.
As a representative of China’s new generation of space-borne long-wavelength Synthetic Aperture Radar (SAR), the LuTan-1A (LT-1A) satellite was launched into a solar synchronous orbit in January 2022. The SAR onboard the LT-1A satellite operates in the L band and exhibits various earth observation capabilities, including single-polarization, linear dual-polarization, compressed dual-polarization, and quad-polarization observation capabilities. Existing research has mainly focused on LT-1A interferometric data acquisition capabilities and the accuracy evaluation of digital elevation models and displacement measurements. Research on the radiometric and polarimetric accuracy of the LT-1A satellite is limited. This article uses tropical rainforest vegetation as a reference to evaluate and analyze the radiometric error and polarimetricstability of the LT-1A satellite in the full polarization observation mode through a self-calibration method that does not rely on artificial calibrators. The experiment demonstrates that the LT-1A satellite has good radiometric stability and polarimetric accuracy, exceeding the recommended specifications of the International Organization for Earth Observations (Committee on Earth Observation Satellites, CEOS). Fluctuations in the Normalized Radar Cross-Section (NRCS) error within 1,000 km of continuous observation are less than 1 dB (3σ), and there are no significant changes in system radiometric errors of less than 0.5 dB (3σ) when observation is resumed within five days. In the full polarization observation mode, the system crosstalk is less than −35 dB, reaching as low as −45 dB. Further, the cross-polarization channel imbalance is better than 0.2 dB and 2°, whilethe co-polarization channel imbalance is better than 0.5 dB and 10°. The equivalent thermal noise ranges from −42～−22 dB, and the average equivalent thermal noise of the system is better than −25 dB. The level of thermal noise may increase to some extent with increasing continuous observation duration. Additionally, this study found that the ionosphere significantly affects the quality of the LT-1A satellite polarization data, with a Faraday rotation angle of approximately 5°, causing a crosstalk of nearly −20 dB. In middle- and low-latitude regions, the Faraday rotation angle commonly ranges from 3° to 20°. The Faraday rotation angle can cause polarimetric distortion errors between channels ranging from −21.16～−8.78 dB. The interference from the atmospheric observation environment is considerably greater than the influence of about −40 dB system crosstalk errors. This research carefully assesses the radiomatric and polarimetric quality of the LT-1A satellite data considering dense vegetation in the Amazon rainforest and provides valuable information to industrial users. Thus, this research holds significant scientific importanceand reference value.
As a representative of China’s new generation of space-borne long-wavelength Synthetic Aperture Radar (SAR), the LuTan-1A (LT-1A) satellite was launched into a solar synchronous orbit in January 2022. The SAR onboard the LT-1A satellite operates in the L band and exhibits various earth observation capabilities, including single-polarization, linear dual-polarization, compressed dual-polarization, and quad-polarization observation capabilities. Existing research has mainly focused on LT-1A interferometric data acquisition capabilities and the accuracy evaluation of digital elevation models and displacement measurements. Research on the radiometric and polarimetric accuracy of the LT-1A satellite is limited. This article uses tropical rainforest vegetation as a reference to evaluate and analyze the radiometric error and polarimetricstability of the LT-1A satellite in the full polarization observation mode through a self-calibration method that does not rely on artificial calibrators. The experiment demonstrates that the LT-1A satellite has good radiometric stability and polarimetric accuracy, exceeding the recommended specifications of the International Organization for Earth Observations (Committee on Earth Observation Satellites, CEOS). Fluctuations in the Normalized Radar Cross-Section (NRCS) error within 1,000 km of continuous observation are less than 1 dB (3σ), and there are no significant changes in system radiometric errors of less than 0.5 dB (3σ) when observation is resumed within five days. In the full polarization observation mode, the system crosstalk is less than −35 dB, reaching as low as −45 dB. Further, the cross-polarization channel imbalance is better than 0.2 dB and 2°, whilethe co-polarization channel imbalance is better than 0.5 dB and 10°. The equivalent thermal noise ranges from −42～−22 dB, and the average equivalent thermal noise of the system is better than −25 dB. The level of thermal noise may increase to some extent with increasing continuous observation duration. Additionally, this study found that the ionosphere significantly affects the quality of the LT-1A satellite polarization data, with a Faraday rotation angle of approximately 5°, causing a crosstalk of nearly −20 dB. In middle- and low-latitude regions, the Faraday rotation angle commonly ranges from 3° to 20°. The Faraday rotation angle can cause polarimetric distortion errors between channels ranging from −21.16～−8.78 dB. The interference from the atmospheric observation environment is considerably greater than the influence of about −40 dB system crosstalk errors. This research carefully assesses the radiomatric and polarimetric quality of the LT-1A satellite data considering dense vegetation in the Amazon rainforest and provides valuable information to industrial users. Thus, this research holds significant scientific importanceand reference value.
This study proposes a computer vision-assisted millimeter wave wireless channel simulation method incorporating the scattering characteristics of human motions. The aim is to rapidly and cost-effectively generate a training dataset for wireless human motion recognition, thereby avoiding the laborious and cost-intensive efforts associated with physical measurements. Specifically, the simulation process includes the following steps. First, the human body is modeled as 35 interconnected ellipsoids using a primitive-based model, and motion data of these ellipsoids are extracted from videos of human motion. A simplified ray tracing method is then used to obtain the channel response for each snapshot of the primitive model during the motion process. Finally, Doppler analysis is performed on the channel responses of the snapshots to obtain the Doppler spectrograms. The Doppler spectrograms obtained from the simulation can be used to train deep neural network for real wireless human motion recognition. This study examines the channel simulation and action recognition results for four common human actions (“walking” “running” “falling” and “sitting down”) in the 60 GHz band. Experimental results indicate that the deep neural network trained with the simulated dataset achieves an average recognition accuracy of 73.0% in real-world wireless motion recognition. Furthermore, he recognition accuracy can be increased to 93.75% via unlabeled transfer learning and fine-tuning with a small amount of actual data.
This study proposes a computer vision-assisted millimeter wave wireless channel simulation method incorporating the scattering characteristics of human motions. The aim is to rapidly and cost-effectively generate a training dataset for wireless human motion recognition, thereby avoiding the laborious and cost-intensive efforts associated with physical measurements. Specifically, the simulation process includes the following steps. First, the human body is modeled as 35 interconnected ellipsoids using a primitive-based model, and motion data of these ellipsoids are extracted from videos of human motion. A simplified ray tracing method is then used to obtain the channel response for each snapshot of the primitive model during the motion process. Finally, Doppler analysis is performed on the channel responses of the snapshots to obtain the Doppler spectrograms. The Doppler spectrograms obtained from the simulation can be used to train deep neural network for real wireless human motion recognition. This study examines the channel simulation and action recognition results for four common human actions (“walking” “running” “falling” and “sitting down”) in the 60 GHz band. Experimental results indicate that the deep neural network trained with the simulated dataset achieves an average recognition accuracy of 73.0% in real-world wireless motion recognition. Furthermore, he recognition accuracy can be increased to 93.75% via unlabeled transfer learning and fine-tuning with a small amount of actual data.
This study focuses on integrating optical and radar sensors for human pose estimation. Based on the physical correspondence between the continuous-time micromotion accumulation and pose increment, a single-channel ultrawideband radar human-pose incremental estimation scheme is proposed. Specifically, by constructing a spatiotemporal incremental estimation network, using spatiotemporal pseudo-3D convolutional and time-domain-dilated convolutional layers to extract spatiotemporal micromotion features step by step, mapping these features to human pose increments within a time period, and combining them with the initial pose values provided by optics, we can realize a 3D pose estimation of the human body. The measured data results show that fusion attitude estimation achieves an estimation error of 5.38 cm in the original action set and can achieve continuous attitude estimation for the period of walking actions. Comparison and ablation experiments with other radar attitude estimation methods demonstrate the advantages of the proposed method.
This study focuses on integrating optical and radar sensors for human pose estimation. Based on the physical correspondence between the continuous-time micromotion accumulation and pose increment, a single-channel ultrawideband radar human-pose incremental estimation scheme is proposed. Specifically, by constructing a spatiotemporal incremental estimation network, using spatiotemporal pseudo-3D convolutional and time-domain-dilated convolutional layers to extract spatiotemporal micromotion features step by step, mapping these features to human pose increments within a time period, and combining them with the initial pose values provided by optics, we can realize a 3D pose estimation of the human body. The measured data results show that fusion attitude estimation achieves an estimation error of 5.38 cm in the original action set and can achieve continuous attitude estimation for the period of walking actions. Comparison and ablation experiments with other radar attitude estimation methods demonstrate the advantages of the proposed method.
,
Through-wall human pose reconstruction and behavior recognition have enormous potential in fields like intelligent security and virtual reality. However, existing methods for through-wall human sensing often fail to adequately model four-Dimensional (4D) spatiotemporal features and overlook the influence of walls on signal quality. To address these issues, this study proposes an innovative architecture for through-wall human sensing using a 4D imaging radar. The core of this approach is the ST2W-AP fusion network, which is designed using a stepwise spatiotemporal separation strategy. This network overcomes the limitations of mainstream deep learning libraries that currently lack 4D convolution capabilities, which hinders the effective use of multiframe three-Dimensional (3D) voxel spatiotemporal domain information. By preserving 3D spatial information and using long-sequence temporal information, the proposed ST2W-AP network considerably enhances the pose estimation and behavior recognition performance. Additionally, to address the influence of walls on signal quality, this paper introduces a deep echo domain compensator that leverages the powerful fitting performance and parallel output characteristics of deep learning, thereby reducing the computational overhead of traditional wall compensation methods. Extensive experimental results demonstrate that compared with the best existing methods, the ST2W-AP network reduces the average joint position error by 33.57% and improves the F1 score for behavior recognition by 0.51%.
Through-wall human pose reconstruction and behavior recognition have enormous potential in fields like intelligent security and virtual reality. However, existing methods for through-wall human sensing often fail to adequately model four-Dimensional (4D) spatiotemporal features and overlook the influence of walls on signal quality. To address these issues, this study proposes an innovative architecture for through-wall human sensing using a 4D imaging radar. The core of this approach is the ST2W-AP fusion network, which is designed using a stepwise spatiotemporal separation strategy. This network overcomes the limitations of mainstream deep learning libraries that currently lack 4D convolution capabilities, which hinders the effective use of multiframe three-Dimensional (3D) voxel spatiotemporal domain information. By preserving 3D spatial information and using long-sequence temporal information, the proposed ST2W-AP network considerably enhances the pose estimation and behavior recognition performance. Additionally, to address the influence of walls on signal quality, this paper introduces a deep echo domain compensator that leverages the powerful fitting performance and parallel output characteristics of deep learning, thereby reducing the computational overhead of traditional wall compensation methods. Extensive experimental results demonstrate that compared with the best existing methods, the ST2W-AP network reduces the average joint position error by 33.57% and improves the F1 score for behavior recognition by 0.51%.
Low-frequency Ultra-WideBand (UWB) radar offers significant advantages in the field of human activity recognition owing to its excellent penetration and resolution. To address the issues of high computational complexity and extensive network parameters in existing action recognition algorithms, this study proposes an efficient and lightweight human activity recognition method using UWB radar based on spatiotemporal point clouds. First, four-dimensional motion data of the human body are collected using UWB radar. A discrete sampling method is then employed to convert the radar images into point cloud representations. Because human activity recognition is a classification problem on time series, this paper combines the PointNet++ network with the Transformer network to propose a lightweight spatiotemporal network. By extracting and analyzing the spatiotemporal features of four-dimensional point clouds, end-to-end human activity recognition is achieved. During the model training process, a multithreshold fusion method is proposed for point cloud data to further enhance the model’s generalization and recognition capabilities. The proposed method is then validated using a public four-dimensional radar imaging dataset and compared with existing methods. The results show that the proposed method achieves a human activity recognition rate of 96.75% while consuming fewer parameters and computational resources, thereby verifying its effectiveness.
Low-frequency Ultra-WideBand (UWB) radar offers significant advantages in the field of human activity recognition owing to its excellent penetration and resolution. To address the issues of high computational complexity and extensive network parameters in existing action recognition algorithms, this study proposes an efficient and lightweight human activity recognition method using UWB radar based on spatiotemporal point clouds. First, four-dimensional motion data of the human body are collected using UWB radar. A discrete sampling method is then employed to convert the radar images into point cloud representations. Because human activity recognition is a classification problem on time series, this paper combines the PointNet++ network with the Transformer network to propose a lightweight spatiotemporal network. By extracting and analyzing the spatiotemporal features of four-dimensional point clouds, end-to-end human activity recognition is achieved. During the model training process, a multithreshold fusion method is proposed for point cloud data to further enhance the model’s generalization and recognition capabilities. The proposed method is then validated using a public four-dimensional radar imaging dataset and compared with existing methods. The results show that the proposed method achieves a human activity recognition rate of 96.75% while consuming fewer parameters and computational resources, thereby verifying its effectiveness.
Recent research on radar-based human activity recognition has typically focused on activities that move toward or away from radar in radial directions. Conventional Doppler-based methods can barely describe the true characteristics of nonradial activities, especially static postures or tangential activities, resulting in a considerable decline in recognition performance. To address this issue, a method for recognizing tangential human postures based on sequential images of a Multiple-Input Multiple-Output (MIMO) radar system is proposed. A time sequence of high-quality images is achieved to describe the structure of the human body and corresponding dynamic changes, where spatial and temporal features are extracted to enhance the recognition performance. First, a Constant False Alarm Rate (CFAR) algorithm is applied to locate the human target. A sliding window along the slow time axis is then utilized to divide the received signal into sequential frames. Next, a fast Fourier transform and the 2D Capon algorithm are performed on each frame to estimate range, pitch angle, and azimuth angle information, which are fused to create a tangential posture image. They are connected to form a time sequence of tangential posture images. To improve image quality, a modified joint multidomain adaptive threshold–based denoising algorithm is applied to improve the image quality by suppressing noises and enhancing human body outline and structure. Finally, a Spatio-Temporal-Convolution Long Short Term Memory (ST-ConvLSTM) network is designed to process the sequential images. In particular, the ConvLSTM cell is used to extract continuous image features by combining convolution operation with the LSTM cell. Moreover, spatial and temporal attention modules are utilized to emphasize intraframe and interframe focus for improving recognition performance. Extensive experiments show that our proposed method can achieve an accuracy rate of 96.9% in classifying eight typical tangential human postures, demonstrating its feasibility and superiority in tangential human posture recognition.
Recent research on radar-based human activity recognition has typically focused on activities that move toward or away from radar in radial directions. Conventional Doppler-based methods can barely describe the true characteristics of nonradial activities, especially static postures or tangential activities, resulting in a considerable decline in recognition performance. To address this issue, a method for recognizing tangential human postures based on sequential images of a Multiple-Input Multiple-Output (MIMO) radar system is proposed. A time sequence of high-quality images is achieved to describe the structure of the human body and corresponding dynamic changes, where spatial and temporal features are extracted to enhance the recognition performance. First, a Constant False Alarm Rate (CFAR) algorithm is applied to locate the human target. A sliding window along the slow time axis is then utilized to divide the received signal into sequential frames. Next, a fast Fourier transform and the 2D Capon algorithm are performed on each frame to estimate range, pitch angle, and azimuth angle information, which are fused to create a tangential posture image. They are connected to form a time sequence of tangential posture images. To improve image quality, a modified joint multidomain adaptive threshold–based denoising algorithm is applied to improve the image quality by suppressing noises and enhancing human body outline and structure. Finally, a Spatio-Temporal-Convolution Long Short Term Memory (ST-ConvLSTM) network is designed to process the sequential images. In particular, the ConvLSTM cell is used to extract continuous image features by combining convolution operation with the LSTM cell. Moreover, spatial and temporal attention modules are utilized to emphasize intraframe and interframe focus for improving recognition performance. Extensive experiments show that our proposed method can achieve an accuracy rate of 96.9% in classifying eight typical tangential human postures, demonstrating its feasibility and superiority in tangential human posture recognition.
Bistatic Synthetic Aperture Radar (BiSAR) needs to suppress ground background clutter when detecting and imaging ground moving targets. However, due to the spatial configuration of BiSAR, the clutter poses a serious space-time nonstationary problem, which deteriorates the clutter suppression performance. Although Space-Time Adaptive Processing based on Sparse Recovery (SR-STAP) can reduce the nonstationary problem by reducing the number of samples, the off-grid dictionary problem will occur during processing, resulting in a decrease in the space-time spectrum estimation effect. Although most of the typical SR-STAP methods have clear mathematical relations and interpretability, they also have some problems, such as improper parameter setting and complicated operation in complex and changeable scenes. To solve the aforementioned problems, a complex neural network based on the Alternating Direction Multiplier Method (ADMM), is proposed for BiSAR space-time adaptive clutter suppression. First, a sparse recovery model of the continuous clutter space-time domain of BiSAR is constructed based on the Atomic Norm Minimization (ANM) to overcome the off-grid problem associated with the traditional discrete dictionary model. Second, ADMM is used to rapidly and iteratively solve the BiSAR clutter spectral sparse recovery model. Third according to the iterative and data flow diagrams, the artificial hyperparameter iterative process is transformed into ANM-ADMM-Net. Then, the normalized root-mean-square-error network loss function is set up and the network model is trained with the obtained data set. Finally, the trained ANM-ADMM-Net architecture is used to quickly process BiSAR echo data, and the space-time spectrum of BiSAR clutter is accurately estimated and efficiently restrained. The effectiveness of this approach is validated through simulations and airborne BiSAR clutter suppression experiments.
Bistatic Synthetic Aperture Radar (BiSAR) needs to suppress ground background clutter when detecting and imaging ground moving targets. However, due to the spatial configuration of BiSAR, the clutter poses a serious space-time nonstationary problem, which deteriorates the clutter suppression performance. Although Space-Time Adaptive Processing based on Sparse Recovery (SR-STAP) can reduce the nonstationary problem by reducing the number of samples, the off-grid dictionary problem will occur during processing, resulting in a decrease in the space-time spectrum estimation effect. Although most of the typical SR-STAP methods have clear mathematical relations and interpretability, they also have some problems, such as improper parameter setting and complicated operation in complex and changeable scenes. To solve the aforementioned problems, a complex neural network based on the Alternating Direction Multiplier Method (ADMM), is proposed for BiSAR space-time adaptive clutter suppression. First, a sparse recovery model of the continuous clutter space-time domain of BiSAR is constructed based on the Atomic Norm Minimization (ANM) to overcome the off-grid problem associated with the traditional discrete dictionary model. Second, ADMM is used to rapidly and iteratively solve the BiSAR clutter spectral sparse recovery model. Third according to the iterative and data flow diagrams, the artificial hyperparameter iterative process is transformed into ANM-ADMM-Net. Then, the normalized root-mean-square-error network loss function is set up and the network model is trained with the obtained data set. Finally, the trained ANM-ADMM-Net architecture is used to quickly process BiSAR echo data, and the space-time spectrum of BiSAR clutter is accurately estimated and efficiently restrained. The effectiveness of this approach is validated through simulations and airborne BiSAR clutter suppression experiments.
微信 | 公众平台 
