This work presents an anomaly detection method based on the Rosenblatt Transformation (RT) for identifying nonhomogeneous range bins in radar data cubes. The algorithm estimates the empirical probability density function from training data and uses the corresponding cumulative distribution function to map a test data vector to independent and identically distributed samples with a uniform distribution. Deviations from uniformity in the transformed variables indicate statistical dissimilarity, allowing the test sample to be flagged as anomalous. We evaluate the method using high-fidelity radar data generated by RFView®, applying the Kullback-Leibler Divergence (KLD) to measure departures from uniformity. A sharp increase in KLD values indicates the onset of nonhomogeneity, consistent with visual observations from range-Doppler and range-angle maps. To validate the method, we estimate the sample covariance matrix using the predicted homogeneous bins and compare it to a ground-truth covariance matrix. Results show that the RT-based approach reliably identifies homogeneous regions and improves covariance estimation in heterogeneous radar environments.
Detecting a target signal in noise is a central problem in radar signal processing. The Normalized Matched Filter (NMF) is widely used, typically evaluated on a discrete grid of target parameters. In practice, targets rarely align with grid points, reducing detection performance. The off-grid NMF addresses this by maximizing over all possible parameters, improving sensitivity to off-grid targets. Building on recent results establishing the relationship between the threshold of the off-grid test and the false alarm probability, we investigate properties of the off-grid NMF that does not seem to be well known in the literature and illustrate through simulations that it exhibits an almost constant false alarm rate in realistic radar scenarios.
Selection of a sequence of pulse repetition frequencies (PRFs) is core to the design of pulse-Doppler radars seeking to detect and decode targets that are ambiguous in range or Doppler. By varying the PRF over multiple coherent processing intervals (CPIs), the radar can decode ambiguous targets and enhance visibility. Pulse sequence selection is constrained by radar system hardware and operating conditions. This paper draws a connection between pulse repetition sequence selection and the problem of finding cliques in a graph. The Bron-Kerbosch algorithm is then shown to efficiently identify mutually compatible sets of PRFs.
Synthetic Aperture Radar (SAR) operating in the P-band enables subsurface imaging and, when combined with the backprojection (BP) algorithm, supports the reconstruction of 3D subtomographic images. To obtain adequate vertical resolution, multi-circular and spiral flight trajectories are employed. As the electromagnetic wave (EMW) propagates through the air-soil interface, refraction takes place, rendering the range calculation non-trivial. This work introduces a modified BP algorithm that incorporates an EMW beam-sampling strategy during the range computation step, accelerated via an Iterative Vectorized Sampling mechanism. Relying exclusively on trigonometric relationships, the method was evaluated using both simulated data of an underground air cavity and real measurements of a buried quad-corner reflector, achieving the expected planimetric and vertical resolutions for both scenarios.
Eye-blink activity provides valuable physiological and behavioral information, yet existing sensing modalities are limited by the need for skin contact or susceptibility to lighting conditions. This paper presents a low-power millimeter-wave multiple-input multiple-output (MIMO) radar-based framework capable of simultaneous, noncontact eye-blink monitoring for multiple subjects located at identical range bins under dynamic clutter conditions. The proposed method is implemented using a 60-GHz frequency-modulated continuous-wave (FMCW) radar platform. The subjects of interest are resolved after calculating the range-angle map for the scene. To extract time–frequency micro-motion features associated with concurrent eyelid movements across subjects, time-frequency analysis is done using short-time Fourier transform (STFT). Experimental results demonstrate reliable and simultaneous eye-blink detection for multiple subjects positioned at the same range, while also highlighting the robustness of the proposed approach against head movements and dynamic environmental clutter. The obtained results demonstrate the feasibility of using radars for applications such as group cognitive-state assessment, multi-person fatigue monitoring in transportation and industrial environments, and simultaneous physiological monitoring for healthcare and human–machine interaction systems.
Reconfigurable intelligent surfaces (RISs) can improve radar target detection by introducing additional controllable propagation paths. In practice, the distributed placement of multiple RISs leads to asynchronous propagation, resulting in path-dependent delay offsets that can degrade the performance of conventional single-matched-filter (SMF) detectors. This work develops a signal model that accounts for such delays and proposes a multi-matched-filter (MMF) detection framework, in which each path is processed by a dedicated MF. The approach also includes joint optimization of the transmit beamformer and RIS phase shifts to maximize detection probability. Analytical detection probability expressions are derived for the asynchronous model and verified via Monte Carlo simulations. Simulations show that both MMF and SMF detectors benefit from RIS deployment compared to systems without RISs. The results further indicate that, as the number of RISs increases, MMF achieves greater detection performance improvements than SMF, particularly under asynchronous conditions.
Radars are an important sensor for pushing autonomy. However the ubiquity of radar is still much less than that of camera. The reason is that it is difficult to tune the radar parameters well for a particular application. In this paper we explore foundation models for Radio Frequency (RF) data. In particular, we show how one can use special self-supervised loss functions to train models for tasks like Direction-of-Arrival estimation (DoA) with super-resolution capability. We propose the first self-supervised model that achieves angular super-resolution in an antenna constrained setting using only a single measurement snapshot. Our model beats the performance of MUSIC algorithm on almost all defined metrics and also achieves angular resolution equivalent to using 8x more antenna elements with the FFT algorithm on real-world datasets.
This paper presents a low-profile, polarization-switchable, circularly-polarized (CP) antenna with an enhanced axial-ratio (AR) bandwidth operating from 8.9 GHz to 10.1 GHz. The AR bandwidth is significantly enhanced from the conventional ~5% to 14.6% by truncating or isolating the four corners of the U-slot square patch. Furthermore, the effective shape of the U-slot is electronically reconfigured using two p-i-n diodes, enabling switching between right-handed CP (RHCP) and left-handed CP (LHCP). The isolated corners conveniently accommodate the biasing circuit required for polarization switching, effectively mitigating its impact on radiation characteristics without increasing antenna size or structural complexity. Simulated and measured results show good agreement, with |S11| below −6 dB and AR under 3 dB across the 8.9 GHz to 10.1 GHz band, yielding a 12.6% overlapping CP bandwidth. The prototype antenna achieves peak gains of 5.36 dBic (RHCP) and 5.24 dBic (LHCP), with a 3-dB-variation gain bandwidth spanning 8.7 GHz to 10.0 GHz. The verified wide CP bandwidth, low-profile, and stable polarization-switching performance make the proposed design highly suitable for X-band pulse radar applications.
Radio-frequency (RF) spectrum is increasingly congested with emissions from radar, communication, and sensing systems across wide bands, demanding advanced spectral awareness for interference management and passive monitoring. This paper presents a blind wideband receiver system for linear frequency-modulated (LFM) chirp detection and estimation without prior knowledge of the waveform. The proposed complex ratio cyclic-difference (CRCD) method exploits quadratic phase coherence, enabling detection without reference and closed-form estimation of chirp rate and center frequency. A Neyman–Pearson framework ensures controlled false-alarm rates, while multi-channel averaging improves robustness under low SNR. The algorithm is validated through MATLAB simulations and executed on an FPGA-based emulation of a heterogeneous system-on-chip for real-time block-wise processing. The results demonstrate accurate parameter estimation and reliable detection, confirming the feasibility of CRCD-based blind chirp sensing for real-time passive spectrum monitoring.
This paper verifies the monostatic and bistatic RCS of a UAV measured in an anechoic chamber. Three different RCS measurements were considered: one monostatic and two bistatic measurements at bistatic angles of 135° and 180°, respectively, for a wide range of aspect angles. Moreover, the measured results were compared with the simulated results obtained by full-wave simulations based on the Shooting Bouncing Ray (SBR) method in CST. Furthermore, the challenges faced during the measurements, such as antenna coupling and meeting the far-field criteria, are described and the methods for overcoming these limitations are provided and discussed.
High-resolution frequency estimation plays a crucial role in radar and spectral sensing applications but remains challenging under low SNR and closely spaced frequencies. Conventional frequency estimation methods suffer from limited resolution and poor noise robustness, while recent deep learning approaches either lack adaptability or require complex architectures. This paper introduces HarmoniNet, a lightweight two-stage neural network that learns a differentiable harmonic projection and reconstructs a fine-grained frequency representation directly from raw complex I/Q data. We propose a new labeling that enhances resolution, and the proposed model jointly optimizes an asymmetric peak-weighted loss and a Gaussian-renderer loss to emphasize weak spectral components and structural accuracy. Experimental results show that HarmoniNet achieves sharper and more accurate frequency localization than existing techniques across SNR levels, successfully resolves closely spaced tones, and extends effectively to radar range profile and Direction of arrival estimation tasks.
With the improvement of radar resolution and the prolonged illumination time, the detection of range-Doppler spread targets (RDSTs) has become increasingly important. However, RDSTs exhibits severe energy dispersion across multiple isolated scattering centers (SCs) with unknown spatial distributions and Doppler frequency centers (DFCs). Conventional detectors are unable to effectively accumulate the dispersed target energy, which could lead to significant performance degradation in RDST detection. To address this challenge, we propose an adaptive detection method for RDSTs based on the joint estimation of SCs and DFCs under the generalized likelihood ratio test criterion (JESCDF-GLRT). In this approach, an augmented target signal vector is formed by stacking the signals from all target SCs, capturing information about both the SCs and DFCs while exhibiting inherent sparsity. A sparse Bayesian optimization model is then formulated to jointly estimate the unknown spatial distribution of the target SCs and the corresponding Doppler steering matrix. By fully exploiting the target energy dispersed across SCs and DFCs, the proposed JESCDF-GLRT method significantly improves RDST detection performance. Its effectiveness is demonstrated through numerical simulations across five target scattering models and benchmarked against competing methods.
We present an all-digital software defined radio (SDR) architecture for time-of-flight (ToF) measurements using the same principles as employed by the global positioning system (GPS). The architecture has been implemented in the register transfer level (RTL) language Verilog and deployed on an AMD Radio-Frequency System on Chip (RFSoC) in loopback mode. We present ToF measurements with accuracy in the order of 10s of nanoseconds. Special treatment is given to the SDR architecture: the phase-locked loop (PLL) implementation for carrier recovery, challenges related to signal processing at data rates above the FPGA fabric speed, the practical implementation of automatic gain control (AGC), as well as design decisions due to resource constraints primarily relating to correlation with the received signal.
Increasingly complex, highly automated, and autonomous driving functions require not only precise, real-time-capable algorithms but also the integration of high-resolution sensor modalities to achieve detailed environmental perception and enable predictive vehicle safety systems. The combination of optical and radar-based sensors opens new possibilities for robust, reliable object detection, but it also presents challenges, particularly under adverse weather conditions such as rain or fog. These environmental factors can significantly impair sensor performance, degrade signal quality, and compromise the reliability of object classification by AI algorithms. This paper investigates the behavior of radar and lidar sensors under controlled, reproducible conditions in the CARISSMA-ISAFE weather facility. First, the effects of varying rain intensity and fog on the performance of a lidar sensor are analyzed. While the primary goal of this work lies in radar sensing, lidar is included as a cross-modal reference with an identical experimental setup. In a second step, the corresponding effects on radar sensor performance are examined. Results from different rain conditions are used to validate weather models under adverse conditions. which can be integrated into simulation tools to produce realistic virtual radar data.
Accurate contactless heart rate (HR) monitoring based on biomedical radar is crucial for providing valuable health information. However, respiratory interference is recognized as a significant challenge, hindering long-term, precise physiological sensing. To address these challenges, a frequency-domain motion reconstruction technique (FDMRT) is proposed, which allows for the linear extraction of the time-domain waveforms of respiratory components based on the frequency information of chest motion. Furthermore, the frequency-domain features of the reconstructed signal and chest motion are processed to facilitate the removal of respiratory, thereby obtaining a clean heart motion spectrum and enabling accurate heart rate estimation. A custom-designed K-band continuous-wave (CW) biomedical radar is developed to implement the proposed technique. Experiments were conducted to validate the proposed method, with accurate heart rate estimation achieved using the radar prototype system. The root mean square error (RMSE) compared to the medical gold standard is found to be 1.87 bpm. The proposed system demonstrates outstanding performance in non-contact vital sign monitoring.
The increasing availability of very high-resolution (VHR) spaceborne Synthetic Aperture Radar (SAR) imagery is creating new opportunities in maritime surveillance, by enabling a more detailed characterization of both primary scattering from the vessel and secondary scattering from its wake. However, accurate kinematic estimation of moving ships remains a significant challenge, as the hybrid SAR/Inverse SAR (ISAR) techniques may exhibit limitations dependent on the target’s kinematics and scattering characteristics. This paper puts forward an integrated framework that complements hybrid SAR/ISAR processing with wake-feature analysis to improve motion estimation reliability. The framework leverages a dual-role strategy: wake-derived parameters are used to independently validate ISAR results when the process is successful, and to provide robust alternative estimates when ISAR fails. Validation on VHR Capella Space data, using two real-world maritime targets with distinct motion characteristics, confirms that this complementary approach yields more robust and reliable kinematic estimates, thereby significantly enhancing maritime situational awareness.
Wind turbines have become an insidious source of clutter contamination in radar observations due to their large radar cross sections (RCS) and their time-varying Doppler signatures from rotating blades. This study investigates the use of adaptive digital beamforming (DBF) with the Capon (also called as Minimum Variance Distortionless Response) method to mitigate wind turbine interference using data collected by Horus, a fully digital phased array radar (PAR) developed at the University of Oklahoma’s Advanced Radar Research Center (ARRC). A field campaign was conducted at the Redbed Wind Farm in Tuttle, OK, where Horus operated in a spotlight mode with a two-dimensional spoiled transmit beam and 52 simultaneous digital subarrays for each polarization. The conventional Fourier DBF and the Capon DBF methods were compared to evaluate their spatial and Doppler-domain performance. Preliminary results show that the Capon method effectively suppresses sidelobe contamination and reduces Doppler contamination from turbine blades, except when the radar beam is directly steered toward the turbines. The adaptive method also demonstrates improved spatial data quality. However, practical limitations include the need for a large number of samples to achieve a full-rank covariance matrix inversion and the computational cost associated the inversion process. Despite these constraints, results indicate that adaptive DBF provides a promising approach for mitigating wind turbine clutter in next-generation weather surveillance radars, improving data quality in clutter-contaminated regions.
In this paper, we investigate spectrally agile electronic protection (EP) countermeasures against knowledge-based (KB) radar jammers. Specifically, we develop interference avoidance transmit-adaptive waveforms along with corresponding receiver matched filtering techniques to mitigate the effects of transmit waveform-shaped noise jammers (TWS-NJ) and its variants. As a form of electronic attack (EA), TWS-NJs generate interference waveforms that are spectrally shaped from the victim transmit signal to degrade radar detection performance and user confidence. Previous works have demonstrated the use of generalized matched filters against known TWS-NJs, but consequentially require significant increased signal-to-noise ratio (SNR) to meet acceptable performance levels as a result of signal energy loss due to main lobe filtering. In this paper, we demonstrate improved EP performance with transmit-adaptive radar waveforms that are spectrally formed to null or avoid TWS-NJ interference when indications and warnings (I&W) are received via electronic surveillance.
This paper addresses the problem of canceling clutter and strong unwanted signals from radar returns for improved detection of low signal-to-noise ratio (SNR) targets. This is especially applicable in Over-the-Horizon Radar (OTHR) where high dynamic range signals are observed. The proposed techniques are based on Thomson’s multitapers – an orthonormal set of discrete prolate spheroidal (Slepian) sequences – formulated in terms of subspace projections. Previously proposed delay Slepians are revisited and a new framework based on combined delay-Doppler Slepians is developed for canceling a specified area in the range-Doppler map with high precision. A computationally feasible method for computing delay-Doppler Slepians for full OTHR waveform is also introduced. New techniques are demonstrated using simulated OTHR data.
Recent advancements in space situational awareness have significantly enhanced the detection and monitoring capabilities of objects in low Earth orbit (LEO). Tracking satellites and space debris is essential for both military and civilian applications. However, the considerable radar-to-target distance significantly hampers the detection of orbital targets, especially with low radar cross-sections. Consequently, advanced signal processing techniques are often necessary to improve detection performance. This paper presents an experimental analysis of how the polarization of target echoes affects detection quality. A novel method for polarization analysis in radar systems is proposed, followed by an evaluation of both simulated and real-world data acquired using a bistatic passive radar. The real signals, including echoes from the International Space Station (ISS) as it passed over Europe, were collected with the Low-Frequency Array (LOFAR) radio telescope. Digital audio broadcasting (DAB+) transmission served as the illuminator of opportunity. The results clearly demonstrate that the polarization state of the echo signal affects the signal-to-noise ratio (SNR), thereby influencing the overall effectiveness of space target detection. The introduced polarization estimation method offers new opportunities for improving the detection and tracking performance of objects in LEO.
