This PDF file contains the front matter associated with SPIE Proceedings Volume 11742 including the Title Page, Copyright information, and Table of Contents.

Introduction to SPIE Defense and Commercial Sensing conference 11742: Radar Sensor Technology XXV

In high density communication and radar environments, radio frequency (RF) signal processing for receivers faces significant challenges. Receivers may receive the overlapped or pulse-on-pulse (POP) RF signals transmitted by cochannel and co-site RF emitters. Detecting, separating, and classifying these POP signals are valuable, but difficult due to the overlapping in time and frequency. In this paper, we propose a time-frequency manifold representation to solve these challenging problems. Using time frequency analysis, we show that a frequency modulation RF (FMRF) signal can be represented as a one dimensional manifold embedded in a two dimensional time frequency space. Using graph theory, we propose a path finding approach to extract this time frequency manifold. With both theoretical analysis and experiments, we show that the proposed approach can extract both simple single and complicated pulse-on-pulse FMRF signals

An algorithm is developed based on Edmund K. Miller’s Model-Based Parameter Estimation (MBPE) technique to mitigate the effects of missing or corrupted data in random regions of wideband linear frequency modulated (LFM) radar signals. Two methods of applying MBPE in the spectral/frequency domain are presented that operate on either the full complex data or separated magnitude/phase data, respectively. The final algorithm iteratively applies MBPE using the latter approach to re-generate results in the corrupted regions of a windowed LFM signal until the difference is minimized relative to un-corrupted data. Several sets of simulations were conducted across many randomized gap parameters where impulse response (IPR) impacts are summarized. Conditions where the algorithm successfully improved the IPR for a single target are provided. The algorithm’s effectiveness on multiple targets, especially when the corrupted regions are relatively large compared to the overall bandwidth of the signal, are also explored.

This paper considers the autofocusing problem for a rotorcraft-borne forward-looking multiple-input multipleoutput (MIMO) synthetic aperture radar (SAR) system. Autofocusing is critically important for forward-looking MIMO SAR imaging, since the radar position errors due to the limited precision of navigation systems can severely degrade the imaging quality. In this paper, we propose a novel autofocusing method to address the radar position errors, resulting from both translational and rotational motion errors, for forward-looking MIMO SAR imaging. The effectiveness and efficiency of the proposed autofocusing algorithm are demonstrated by using several numerical experiments.

In radar target tracking, knowledge of the true dynamics of target motion is paramount for accurate state estimates. In this paper, we propose a method of target maneuver detection utilizing symbolic dynamics. We demonstrate its ability to compete with other commonly used maneuver detectors. This is done through simulations performing target maneuver detection.

At present, the majority of deep learning building blocks, techniques, and architectures are based on real-valued operations and representations. However, over the past decades, neural networks in the complex domain have become a quite active field of research and have continued to open doors to various new applications, remote sensing being one of them. Complex-valued neural networks (CVNNs) deal with complex inputs, e.g. signals having phase and amplitude, which is one of the core concepts in signal processing. Despite their appealing properties and potential for opening up entirely new neural architectures, deep complex-valued neural networks have been marginalized due to limited availability of building blocks required for such model designing. This survey compactly summarizes the research efforts that demonstrates relevant applications of deep complex-valued neural networks in the realm of remote sensing. We examine the various remote sensing problems under study, the models employed and the sources of data used. An attempt is made to study the overall performance achieved by such complex-valued models, according to the evaluation metrics used by the authors and compare the competencies with their real-valued counterparts. Also, we illustrate the shortcomings, enhancements, and implementations associated and obtain a bird's-eye view of their present and future prospective. The overall findings indicate that CVNNs represent a promising technique with competitive performance in terms of classification accuracy and precision for a wide variety of remote sensing problems, outperforming conventional techniques. However, the success of each CVNN model is highly dependent on the nature of the data set used.

Deep convolutional neural networks (CNNs) provide the sensing and detection community with a discriminative, machine learning based approach for classifying images of objects. However, one of the largest limitations for deep CNN image classifiers is the need for extensive training data for a variety of appearances of class objects. While current methods such as GAN data augmentation, noise-perturbation, and rotation or translation of images can allow CNNs to better associate convolved features to ones similar to a learned image class, many fail to provide new context of ground truth information associated with each object class. To expand the association of new convolved feature examples with image classes within CNN training datasets, we propose a feature learning and training data enhancement paradigm via a multi-sensor domain data augmentation algorithm. This algorithm uses a mutual information, merit-based feature selection subroutine to iteratively select SAR object features that most correlate to each sensor domain’s class image objects. It then re-augments these features into the opposite sensor domain’s feature set via a highest mutual information, cross sensor domain image concatenation function. This augmented set then acts to retrain the CNN to recognize new cross domain class object features that each respective sensor domain’s network was not previously exposed to. Our experimental results using T60- class vs T70-class SAR object images from both the MSTAR and MGTD dataset repositories demonstrated an increase in classification accuracy from 88% and 61% to post-augmented cross-domain dataset training of 93.75% accuracy for the MSTAR, MGTD and subsequent fused datasets, respectively.

Internet of Things (IoT) has become a fast growing research topic in recent years. Internet connected sensors and devices allow for the collection and processing of a wealth of data. This in conjunction with sensor fusion can provide greater accuracy in object recognition and detection surpassing what could be obtained by sensors operating independently. However these distributed sensors must often operate in environments with poor to no access to the internet which can greatly reduce their effectiveness. Additionally these sensors can be attached to highly dynamic platforms further complicating communication and data routing. One possible solution is to use the B.A.T.M.A.N. (Better Approach To Mobile Ad hoc Networking) routing protocol adapted for use with LoRa, a low power long range RF protocol, to route sensor data through other nodes in order to reach internet access points and allow these devices to interact with the cloud that would have otherwise been unable. Other adaptations to the algorithm will be investigated, such as including other sensors, like GPS and message signal strength to better predict route quality. This system shows promise to be an effective, fault-tolerant solution for this application.

Implementation of a real-time radar application for a software defined radio (SDR) application presents serious challenges to the software engineer. All of the critical tasks, such as pulse compression and target detection, must be performed within a specified time interval. The efficiency of this implementation determines the amount of downstream processing that can be performed while adhering to strict timing requirements. Researchers at ARL have implemented SDR-based radar applications in the past; however, they have never described the software engineering required to implement such a system. This paper documents, in some detail, the issues associated with implementation of real-time radar processing code. It begins with a bird’s eye view of the software functions, outlined using high-level block diagrams. It then touches upon some of the lower-level coding issues associated with realization of the high-level functionality.

Frequency modulated continuous wave (FMCW) radar allows for a wide range of research applications. One primary use of this technology which is explored in this paper is the ground penetrating radar. To achieve high sensing performance, wide-band spectral reconstruction and sophisticated image reconstruction algorithm have been developed to overcome hardware limitations. Applications and future work include Synthetic Aperture Radar (SAR) imaging, innovative GPR, and unmanned aerial vehicle (UAV) radar systems.

Using ultra-wideband (UWB) impulse radar for detecting and tracking fast-moving small targets over the ocean surface has been considered before with limited applications. The challenges of deploying such radar sensors on small, unmanned marine platforms are addressed in this study. The first challenge is the stringent size and weight requirement to allow a tracking radar sensor to be fitted into the payload of a small unmanned surface vehicle (USV). For the first time, we implemented a design that is based on a single chip UWB radar sensor operating at X-band, which effectively achieves the size and weight requirement for a small USV payload. The second challenge is range extension and range ambiguity resolution. With the UWB radar operating various high-PRF modes, we developed a novel approach that stitches together range profiles from multiple PRFs, to extend the effective non-ambiguous range at the cost of scan speed. The third challenge is developing a lowcost, ultra-wideband planar antenna and front-end, which is also part of the USV payload and needs to be able to perform either sector scanning, or even electronic scanning, with a very low profile. We have successfully designed and implemented one such antenna using a dipole array design. By integrating the solutions into a complete system, we have performed a series of lab and outdoor tests of the UWB radar sensor and obtained some promising target data. Simulations are also being developed for testing the potential target signatures and tracking effectiveness of moving targets over ocean surface clutter environments.

In recent years, quantum radar has focused entirely on using bipartite squeezed states of light as a mechanism for target detection. This paper studies the performance of a quantum radar that uses a tripartite squeezed state, whereby two signal beams are sent out towards the target which both correlate with the idler. It is found that for very low signal strengths, the bipartite has better performance. As the signal strength increases however, the tripartite becomes dominant. This result suggests that quantum radar (declared useful only in the low SNR regime) may possess more possibilities of increased performance at higher SNRs when different states are used for correlation. The bottleneck, of course, is the ability to generate transmit powers necessary to utilize.

Intermodulation radar is an established technique for locating electromagnetically-nonlinear junctions. For this type of radar, the probe consists of multiple simultaneous frequencies, usually two tones of equal amplitude. The multiple frequencies illuminate the target, mix with each other, and generate integer sums and differences of the original transmitted tones. This work studies a variation on the intermodulation-radar technique. Some targets, such as AM/FM transmitters, emit radio frequencies without being actively probed; thus, some collections of (powered) nonlinear junctions generate at least one internal tone which might be mixed with an externally-applied probe tone. This internal-external mixing is referred to as “carrier modulation,” where the carrier is associated with the target and its modulation is induced by the transmit probe. This paper documents an experiment conducted using a transverse electromagnetic cell: contactless excitation of carrier modulation from active nonlinear junctions. Data recorded from two radio transmitters indicate that, for this internal-external mixing technique, a reduction in transmit power results in less of a reduction in received power compared to traditional intermodulation radar.

Radar waveform design is a long standing area of interest and research in the radar community. Recently, radar waveform design has been paired with Sudoku, a logic puzzle that has been popular worldwide since the 1980s. The combination of the two has spurred a new area of radar research: Sudoku based radar waveform design. In this paper, we compare Sudoku based Multicarrier Phase-Coded (MCPC) signals to both Costas codes and random codes for different transmitted pulse structures. We provide computational results in the form of statistical analysis for the autocorrelation function and analytical ambiguity function analysis, as well as measurements of the Peak Side Lobe Level (PSL) of their Autocorrelation Function (ACF). We then assess the closeness of the resulting ambiguity function to the ideal thumbtack response.

One obtains a frequency-hopping waveform from a traditional 9 × 9 Sudoku by zeroing out all entries besides a fixed integer and using the non-zero integer positions as the frequency-hopping permutation. Using a new and general approach to describing Doppler tolerance of Setlur et al., we previously computed the narrowband ambiguity function to formally compare Doppler tolerances of Sudoku-based frequency-hopping waveforms with Doppler tolerances of waveforms made from Costas arrays and from random permutations. The analyses in this paper will expand our previous analyses to pulse trains, to Sudokus having additional constraints (Hyper Sudoku and X-Sudoku) in support of more waveform structure, and to larger Sudoku (Superdoku) in support of higher time-bandwidth products. The ambiguity analyses investigate the peak sidelobe levels found from using various permutations as well as the sidelobe levels around the central peak of the ambiguity function.

In several wireless applications involving communication and radar systems, the concept of Frequency Diverse Arrays (FDA) is becoming a popular choice. These types of arrays provide greater flexibility than the conventional phased arrays in beam steering and beamforming applications. Even though phased arrays have shown success in beam steering and beamforming with respect to flexible beam scanning in detecting and tracking weak targets, the major drawback is the beam steering occurring at a fixed angle for all ranges. To mitigate this, FDA utilizes applying a phase progression across all the elements in the array. As a result, this enables more degrees of freedom with respect to beam scanning and reducing effects due to multi-path and other interferences. The focus of this paper is to detail the design of FDAs using Sudoku arrays in radar applications involving beamforming and beam steering along with analyzing multiple beams simultaneously forming at different directions. Simulations were performed using MATLAB, in which antenna arrays were designed with uniform spacing of half-wavelength and operated at a constant frequency. Radiation pattern and polar plot of the antenna array were analyzed with respect to sidelobes levels and beamwidth.

The future battlespace requires seamless operation across multiple domains and coordination at echelon to facilitate windows of opportunity for exploitation and joint engagement. Future radar and sensor systems challenged to operate in these dynamic and Multi-Domain environments must overcome the enduring challenges of survivability, redundancy, frequency exclusivity, and GPS-dependence, among others. This drives a need for foundational research in device technology and sensor architectures that provide a path to meeting the long-term vision of next-generation RF sensors that are capable of coordinating in battle with legacy and cross-service assets. This keynote talk, given by the Army’s ST for Electromagnetics, provides a perspective on research initiatives needed to develop next generation capabilities that are GPS-independent, autonomous, low-SWAP, and provide multi-function capability. This provides context for a special session on next-generation electronic materials and ultrawide bandgap semiconductors that enable RF and power devices to provide leap-ahead capabilities in output power and thermal properties. This enables future systems with more power on target capabilities, improved range performance, and lower SWAP and power consumption through improved efficiency and reduced cooling requirements.

Surface induced transfer doping (SITD) is a novel, highly efficiency doping technique that is being used to invoke the p-type surface conductivity of intrinsic diamond for high-frequency, high-power electronic devices. In the SITD process, a high electron affinity (EA) thin film acceptor layer is interfaced with the hydrogenated diamond surface with negative electron affinity (NEA) to induce the effective p-type doping on the diamond surface. Overall, device performance of the SITD doped devices is contingent on the type and quality of the interface between the acceptor layer and hydrogenated diamond surfaces. Motivated by this, our internal theoretical modeling efforts based on a hybrid approach of machine learning and first principle calculations have focused on performing bottom-up design of novel acceptor layers with higher stability and improved device performance, e.g., doped TMOs and 2D layer. In this talk, recent results from our predictive modeling effort will be presented.

High power radio frequency (RF) transfer-doped diamond field effect transistors (FETs) are being fabricated at the Army Research Laboratory (ARL). To implement these into radar systems we have a parallel effort to extract accurate compact models from their measured DC and RF data. At this early stage we are using the commercially available Angelov model and will discuss fitting the model parameters and how their parameter values differ from GaN and GaAs FETs. Results indicate good model prediction of measured results in some cases. Also, model extraction can indicate areas of the device that needs greater attention for improved performance such as the access region resistance. Furthermore, in the saturation region of operation these transistors exhibit a hole saturation velocity of 5 × 10⁶ cm/s obtained from extracted model parameters.

The gate-controlled hole gas at the hydrogenated diamond surface was predicted to have a plasmonic response to a terahertz and sub-terahertz electric field, making p-diamond field effect transistors (FETs) promising candidates for implementing room temperature plasmonic devices. The predicted performance of diamond plasmonic detectors shows their potential for high temperature, high voltage, and radiation hard applications and for THz communications and spectroscopy in the atmospheric windows from 0.2 to 0.6 THz. This makes p-diamond a unique material for Beyond 5G THz communications, since a resonant plasmonic response makes also possible the realization of p-diamond based emitters in sub-terahertz range, using strong current driven plasma instability in gated channels. Toward the optimal design of p-diamond plasmonic devices we simulated the response using hydrodynamic equations, Our fabrication process for obtaining higher mobility p-diamond plasmonic FETs will be presented.

Army Research Laboratory (ARL) is developing radio frequency (RF) field-effect-transistors (FETs) on hydrogen-terminated, single-crystal diamond surfaces. By employing advanced fabrication methods, we achieve state-of-the-art device performance with gate lengths below 100 nm. We are exploring methods to improve the stability of fabricated FETs, which is critical for maturation of the technology and its commercial acceptance. DC and RF measurement data will be reviewed and discussed within the framework of improving device yield and reliability.

We present uncooled, fiber-coupled InGaAs photodiode modules for DC to 18 GHz applications that generate broadband linear RF signal with peak-to-peak output amplitude up to 4 Vpp and phase nonlinearity, i.e. power-to-phase conversion factor <6 rad/W. The photodiode’s phase nonlinearity approaches zero for selected operating conditions, as needed for generating precision clocks. The photodiodes demonstrate reliable operation up to a DC photocurrent of 50 mA at 9 V reverse bias at ambient room temperature of 20 °C with passive heat sinking. The photodiodes are packaged in miniature, fiber-pigtailed modules having a size, weight, and power (SWaP) that is suitable for phased array systems.

Beta-Ga₂O₃ has emerged as a promising ultra-wide-bandgap semiconductor system for next generation power, and RF applications. In this talk, we will present our recent results on gallium oxide based high breakdown voltage MOSFETs and scaled gate length RF devices. The large bandgap of Ga₂O₃ leads to a high critical field strength. This high field strength in combination with high room temperature mobility and calculated electron velocity leads to a higher Baliga’s Figure of Merit (BFoM) and Johnston Figure of Merit (JFoM) than commercial widebandgap technologies. Additionally, the large bandgap also enables high temperature operation and radiation hardness making it attractive for harsh environment applications. We will present our work on the lateral MOSFETs with improved field plate design which show record high breakdown voltages. DC and pulsed I-V characteristics of 100 nm gate length Ga₂O₃ MOSFET will be reported. DC-RF dispersion and passivation technology will be discussed.

Ga₂O₃ is the only ultra-wide bandgap semiconductor with melt-growth substrate technology similar to that of Si, heterostructure device technology similar to that of the III-Nitride family, and high growth rate (GR) epitaxial technologies such as MOCVD and HVPE to support the development of ultra-high-breakdown voltage devices competitive with SiC technology. We report a Ga2O3 transistor device based on a high-GR MOCVD technology (Agnitron Technology’s Agilis 100 reactor). We have demonstrated for the first time a β-Ga₂O₃ MOSFET grown by high-GR MOCVD resulting in significantly improved epilayer quality. The high GR demonstrated via this method paves the road for demonstration of high breakdown voltage devices on a thick Ga₂O₃ buffer layer.

The demand of higher power density from power converters has increased since its very origin and increased by 3 orders of magnitude over the last 40 years. This trend is not just seen in power conversion, but also in Radar communication where RF power density is the metric. So, although very different in applications, both power converter and RF power amplifiers are relying on a roadmap decorated with WBG and Ultra-WBG materials. Beyond GaN and SiC, Ga2O3, Diamond and AlN have been explored for some time in academia, government labs, and industry. All these materials come with their own merits and demerits; doping efficiency being of utmost relevance for their application as a semiconductor device. Diamond, with a bandgap of 5.47eV is definitely a very attractive material for its large critical electric field (3x more than SiC), high thermal conductivity (6x more than SiC) and other carrier transport properties.

High Ice Water Content (HIWC) is an atmospheric condition at high altitude that may lead to failure of jet engines. As a potential threat to aviation safety and space launch operation, it has received significant attentions from cross-disciplinary communities. Detecting HIWC conditions with airborne radar is essential to the safe monitoring of this type of hazard, however it has unique and significant challenges. For example, in general, small ice particles and clusters of ice particles do not register strong radar reflectivity, which is a challenge to the sensitivities and resolutions of small aperture airborne radars. Second, it is difficult to discriminate HIWC from other atmosphere conditions, such as general precipitations, and evaluate the threat level (in quantity of Ice Water Content, or IWC) with remote sensing only. In this study, we developed a novel simulation-based approach, which uses the in-situ probe collected HIWC cloud probe data during a series of flight test campaigns, as well as the microphysical particle models retrieved from these data as the basis of simulations. Then, we combine and reconcile these models with the ground-radar measurements, which leads to a three-dimensional truth gird. Using this truth field, we developed a single-cell-Monte-Carlo (SCMC) simulation implementation, which creates and generates airborne weather radar signatures and moments for each individual resolution cell. The simulation has incorporated (1) An initial framework of airborne radar system and sensor modeling, (2) Modeling of ground clutters and effect of antenna patterns. The simulation tool has significant applications in the areas of (1) Guidance of designing and development of next generation airborne hazard sensing and avoidance radars. (2) Support industry standard making and performance evaluations such as FAA, and (3) Support scientific studies on airborne radar signatures and techniques for further understanding of hazardous atmosphere conditions for aviation.

Synthetic aperture radar (SAR) echo simulation offers a low-cost and convenient way to obtain high-resolution images of targets, and plays an important role in system design and algorithm validation. Although high frequency approximation simulation is widely used, it is considered to be imprecise when calculating scattering field of fine structures, such as exhaust pipes and groove structures, especially in low frequency band. In this paper, a finite-difference time-domain (FDTD) based method is proposed for high-precision SAR echo simulation. In this method, scattering process of electromagnetic wave is accurately simulated to obtain equivalent electric and magnetic current on the surface of the target. Also, a near-to-far-field transformation is applied to the equivalent electric and magnetic current to calculate the field at the receiving antenna. In this transformation, a waveform forming method is introduced to simulate stripmap SAR echoes. By introducing this method, the usage of FDTD in one single simulation can be greatly reduced. The experiments show that proposed method can significantly improve the efficiency of the simulation while maintaining echo accuracy.

There are various scenarios, whether they be commercial or defense, where privacy is important. In communications, the metrics of low probability of interception is often used to measure the signal’s ability to resist interception and decoding by unauthorized parties. Joint radar sensing and communications (RadarCom) has been of interest recently and an important requirement of RadarCom signals is its immunity to interceptions. In this context it is of interest to understand the statistics of background clutter. This paper uses machine learning (ML) approaches to classify and model clutter in presence of noise/interference. We employ 32 sub-carrier orthogonal frequency division multiplexing waveforms as a basis for clutter return collection and subsequent use as RadarCom signals. We then present the ML combination method with the best classification accuracy of 78.9%.

Generation of moving target (Doppler and micro-Doppler) radar data on a scale-model helicopter with an arbitrary flight path is described. Fully-polarimetric micro-Doppler radar signatures of a 1/35th scale helicopter at S-Band were measured using a 100GHz compact range and are examined under several different situations. The motion of the rotor and blades are analyzed using standard range-doppler spectrograms. In particular, the effect of radial motion toward or away from the radar is considered and the consequences this motion will have on the range/doppler spectrogram of the radar data. It is found that a helicopter exhibiting no radial velocity will experience a degeneracy of signals from the rotation of the helicopter blades. This degeneracy is lifted when a non-zero radial motion is experienced. The effect of varying the radar pulse repetition frequency will be examined.

Birds and UAVs have similar Radar Cross Section and speed, making birds confusers when trying to detect and track UAVs. Birds can have micro-Doppler features in their signatures, but collecting bird data at multiple angles and directions can be expensive. An alternative is using micro-Doppler simulations of birds. Accurate simulations are made of a single bird, but can also be used to produce flocks of birds as potential confusers. The bird simulation was analyzed for its similarity to drone or UAV rotor radar micro-Doppler. Using these simulations, the effect of the flight angle relative to the radar beam was analyzed for the amount of observable micro-Doppler, allowing the analysis of track angles and the enhanced probability of a false alarm from a bird. This paper shows the differences in bird micro-Doppler returns simulated over angles and compares them to a micro-Doppler model of a UAV rotor.

Tracking the motion of held objects is becoming increasingly important in smart home applications. MicroDoppler radar has proven beneficial for tracking the motion of people, as various parts of the body, such as the arms and legs, move with different velocities as a function of time, thereby generating different Doppler frequency sidebands in the received response. The time-frequency signature of these responses can be used to classify activities. Since natural objects are generally linear, the back-scattered signals are collected at the same frequency as the transmitted signals, thus the micro-Doppler frequency sidebands are observed around the transmitted center frequency. In home settings, clutter can thus become a challenge in the detection of small movements. In this work, we demonstrate an approach for tracking held objects in high-clutter environments using harmonic Doppler and harmonic tags to detect the micro-motion signatures of held objects. While previous works have investigated harmonic radar for target tracking, this work uniquely focuses on detection of the timevarying Doppler responses from micro-motion signatures of held objects. By placing a passive harmonic tag on various parts of the human body, the motion of individual body parts and/or individual held objects can be discerned. In this work we characterize the harmonic micro-Doppler signatures of tags held on different parts of the body. We present expected results and compare them to measurements conducted using a 2.51 GHz/5.02 GHz harmonic Doppler radar

Most research in technologies for the Deaf community have focused on translation using either video or wearable devices. Sensor-augmented gloves have been reported to yield higher gesture recognition rates than camera-based systems; however, they cannot capture information expressed through head and body movement. Gloves are also intrusive and inhibit users in their pursuit of normal daily life, while cameras can raise concerns over privacy and are ineffective in the dark. In contrast, RF sensors are non-contact, non-invasive and do not reveal private information even if hacked. Although RF sensors are unable to measure facial expressions or hand shapes, which would be required for complete translation, this paper aims to exploit near real-time ASL recognition using RF sensors for the design of smart Deaf spaces. In this way, we hope to enable the Deaf community to benefit from advances in technologies that could generate tangible improvements in their quality of life. More specifically, this paper investigates near real-time implementation of machine learning and deep learning architectures for the purpose of sequential ASL signing recognition. We utilize a 60 GHz RF sensor which transmits a frequency modulation continuous wave (FMWC waveform). RF sensors can acquire a unique source of information that is inaccessible to optical or wearable devices: namely, a visual representation of the kinematic patterns of motion via the micro-Doppler signature. Micro-Doppler refers to frequency modulations that appear about the central Doppler shift, which are caused by rotational or vibrational motions that deviate from principle translational motion. In prior work, we showed that fractal complexity computed from RF data could be used to discriminate signing from daily activities and that RF data could reveal linguistic properties, such as coarticulation. We have also shown that machine learning can be used to discriminate with 99% accuracy the signing of native Deaf ASL users from that of copysigning (or imitation signing) by hearing individuals. Therefore, imitation signing data is not effective for directly training deep models. But, adversarial learning can be used to transform imitation signing to resemble native signing, or, alternatively, physics-aware generative models can be used to synthesize ASL micro-Doppler signatures for training deep neural networks. With such approaches, we have achieved over 90% recognition accuracy of 20 ASL signs. In natural environments, however, near real-time implementations of classification algorithms are required, as well as an ability to process data streams in a continuous and sequential fashion. In this work, we focus on extensions of our prior work towards this aim, and compare the efficacy of various approaches for embedding deep neural networks (DNNs) on platforms such as a Raspberry Pi or Jetson board. We examine methods for optimizing the size and computational complexity of DNNs for embedded micro-Doppler analysis, methods for network compression, and their resulting sequential ASL recognition performance.

Airborne ash generated by explosive volcanic eruptions presents a significant danger to aviation. Accurate modelling and predictions of the dispersal of hazardous ash into the atmosphere are currently hampered by uncertainties in the ‘source term’ parameters associated with the initial eruption plume, specifically the amount and size of ash particles released into the atmosphere. Ground based radar offers the means to remotely measure ash reflectivity, however estimation of source term parameters from reflectivity measured by single frequency radar is limited by ambiguity between the contribution of particle size distribution (PSD) and ash concentration in the plume. This means that one of these parameters must be assumed rather than measured directly, leading to uncertainties in forecasting eruption hazards. We report on R4AsH, a close range FMCW radar designed to resolve this ambiguity by simultaneous characterization of falling volcanic ash in a laboratory-controlled environment at three different frequencies: 10, 35 and 94 GHz. The R4AsH design uses a single DDS based chirp generator as a common source, multiplied and upconverted to feed three sets of transmit-receive horn antennas directed at a common target volume such that measurements will give spatially and temporally coincident measurements of falling ash. In addition, there will be independent measurement of the PSD using optical imaging and logging of the landing particle mass to calibrate results and inform analysis. The aim of R4AsH is to develop a triplefrequency inversion algorithm to enable simultaneous retrieval of PSD and ash concentration from radar data suitable for future volcano monitoring systems.

The nonlinear properties of electronic devices, such as diodes and transistors, can generate nonlinear responses such as harmonic and intermodulation for clutter rejection purposes. The naturally occurring clutter objects are generally linear, or they possess very small nonlinearity. They can be distinguished from man-made targets containing the above-mentioned nonlinear devices by exploiting nonlinear responses. In this paper, this nonlinear property was utilized for generating a 3 rd -order harmonic response for clutter rejection purposes. Existing nonlinear radars generally exploit 2 nd -order harmonic responses for clutter rejection purposes owing to their high power levels among the harmonic responses. However, due to their proximity to the fundamental tone, these radars require bulky and expensive filters and diplexers with steep roll-off for maintaining the linearity of the transmitter and receiver section of the radar. As the 3rd -order harmonic response and the fundamental tone are widely separated in the frequency spectrum compared to the 2nd order harmonic response, better isolation between the fundamental and harmonic response can be achieved. This results in relaxed requirements for filters and diplexers for these 3rd order based harmonic radars. Apart from that, the receiver and tag size would be reduced since size and frequency are inversely proportional. In this paper, the 3 rd -order harmonic radar and a passive tag were designed and operated in millimeterwave frequency bands, i.e., 24/72 GHz. Experimental validations were performed to prove their clutter rejection ability.

Millimeter-wave (MMW) imaging systems require high resolution and spurious free dynamic range (SFDR) to produce images with low artifacts. In these systems, lateral resolution is determined by the center frequency and the ratio of the aperture size to the distance to the target. The downrange resolution is determined by the system bandwidth; for very fine downrange resolution of a target, or material layers/discontinuities, the bandwidth must be very broad. Additionally, the response over the bandwidth must be measured very rapidly to support high speed acquisition over large apertures. A voltage-controlled oscillator (VCO) is a standard means of generating the ultra-broadband frequency chirp. This paper investigates the use of a broadband VCO for use in MMW imaging systems, a linear-phase calibration technique, and digital-to-analog converter (DAC) parameter considerations for controlling the VCO.

Once Synthetic Aperture Radar (SAR) images are formed, they typically need to be stored in some file format which might restrict the dynamic range of what can be represented. Thereafter, for exploitation by human observers, the images might need to be displayed in a manner to reveal the subtle scene reflectivity characteristics the observer seeks, which generally requires further manipulation of dynamic range. Proper image scaling, for both storage and for display, to maximize the perceived dynamic range of interest to an observer depends on many factors, and an understanding of underlying data characteristics. While SAR images are typically rendered with grayscale, or at least monochromatic intensity variations, color might also be usefully employed in some cases. We analyze these and other issues pertaining to SAR image scaling, dynamic range, radiometric calibration, and display.

Often a crucial exploitation of a Synthetic Aperture Radar (SAR) image requires accurate and precise knowledge of its geolocation, or at least the geolocation of a feature of interest in the image. However, SAR, like all radar modes of operation, makes its measurements relative to its own location or position. Consequently, it is crucial to understand how the radar’s own position and motion impacts the ability to geolocate a feature in the SAR image. Furthermore, accuracy and precision of navigation aids like GPS directly impact the goodness of the geolocation solution.

In order to characterize transmissions from a specific radar system, a surveillance team must first detect and then characterize the received radar pulses. Hence, pulse detection and pulse width estimation represent two critical functions of the surveillance activity. We present three procedures for efficiently performing both pulse detection and pulse width estimation. The first procedure leverages an edge detector based on a Haar wavelet filter, the second exploits the ratio of leading and trailing moving averages, and the third incorporates a novel formulation of the ratio that simplifies calculation of the detection threshold. The design features of the algorithms that engender computational efficiency are outlined, and effective methods for determining detection thresholds are described. The techniques are compared and contrasted using measured data from the CCDC U.S Army Research Laboratory’s Advanced Electronic Warfare Lab. Favorable results are obtained for signal-to-noise ratios of approximately 1.7 and 3.0 dB.

Ultra-wideband (UWB) ground-penetrating radar (GPR) technology has been widely employed for detecting targets that range from buried explosive devices such as landmines, improvised explosive devices (IEDs), to underground utilities and tunnels, etc. However, the backscatter signals from the ground surface pose a critical challenge for downward-looking GPR systems since (i) these ground return signals have significant power compared to the backscatter signal from subsurface targets, and (ii) the ground return and target signals completely overlap in both the time and frequency domains. Many techniques have been proposed to date; however, they all have limitations in mitigating the adverse effects of the very high power ground return interference (GRI) signals. This paper presents a novel technique for reconstructing and extracting the GRI signals from downward-looking UWB GPR signals. Our proposed technique performs an estimation of the return signal from the ground surface. This signal estimation, together with the estimated scatter center of the ground surface, is used to construct a dictionary that represents the ground return signal subspace. Finally, we employ a sparsity-driven optimization algorithm to reconstruct the GRI signals and then extract them from the received radar signals. All information used to construct the dictionary is completely derived from the data. Our technique performs this GRI extraction directly in the phase history data domain prior to synthetic aperture radar (SAR) image formation. Thus, it can be implemented as an additional step, completely independent from all other steps, in the pre-processing stage. Recovery results from simulated data set illustrate the robustness and effectiveness of our proposed technique.