Search

In this paper, we employ a 77GHz frequency modulated continuous wave (FMCW) multiple-input-multiple-output (MIMO) radar to achieve hand gesture recognition based on a three-branch convolutional neural network (CNN) with fine-tuning. Firstly, the FMCW MIMO radar is utilized to capture the hand gesture data that are the fast-time-slow-time-antenna 3 dimension (3D) data. Then, by applying the discretize Fourier transform (DFT) to the fast-time and slow-time, the multiple signal classification (MUSIC) approach to the antenna dimension, and the multi-frame accumulation, the range-Doppler-angle temporal signatures can be extracted from the captured data. In order to exploit the temporal and spatial correlations of the hand motions, we construct a three-branch CNN to self-learn the hand gesture signatures from the range, Doppler and angle dimension, respectively, and to recognize 9 hand gestures. The fine-tuning approach is proposed to improve the robustness of the proposed network in recognizing an untrained person's hand gestures. After the fine-tuning, the experiment results show that the proposed approach can recognize 9 hand gestures of all trained persons with an average accuracy over 96% and an untrained person with an average accuracy over 96%.

Radar sensors provide unique advantages in human motion recognition, such as privacy protection, operating free from weather and light, etc. However, it is difficult to obtain large radar samples in contrast to optical images due to the limit of time cost and human resources. Although deep convolutional neural networks (DCNN) have achieved great success in image classification, the lack of training samples remains a key problem. In this paper, we utilize micro-doppler simulated data to enlarge training samples, with which bistatic convolutional neural networks (Bistatic-CNN) are pretrained and later we use transfer learning to improve the accuracy of human motion recognition. Furthermore, we compare online dataset from Carnegie Mellon University (CMU) with Kinect sensor in order to figure out the differences. Finally, both simulated data are used as training samples for transfer learning

Radar-based human activity classification has attracted more and more attention in the fields of anti-terrorism maintenance, post-disaster search and rescue, medical monitoring and smart homes, which not only has the advantage of detecting in harsh environments but also can protect personal privacy to the greatest extent. The combination of deep learning methods and micro-Doppler spectrograms for radar human activity recognition is becoming more and more popular, but the general network model has many parameters and long training time, making it difficult to train a better network model. Therefore, we have designed an efficient and lightweight network model, which is based on the Mobilenetv2 model pre-trained on the ImageNet dataset. The model we proposed can learn the micro-doppler information and can automatically identify tasks of different types of human movements. We use the micro-Doppler spectrogram as the training data of the network model. The model has a small number of parameters. It can train a model very quickly. At the same time, it can trace-off the accuracy and resource requirements. It can be run on the embedded in the system, it can meet the requirements of the human activity recognition system. We have shown very good performance on the experimental data provided by the official, which can not only ensure the accuracy of the model to predict human activities, but also save a lot of training time. At the same time, our experiments found that a single human actvitiy can be quickly and accurately recognized, however, the complex human behavior composed of a series of single actions still needs to continue to work

The marine radars are widely used in surveillance system for its convenience and cost-effectiveness, but it always suffers from the sea clutters and other inferences, and there are multiple maneuvering targets in the surveillance area. In these complex scenes, the traditional radar detector and tracker can result high false alarm and target loss. Especially the high false alarm scenes can bring an unacceptable computational burden. In this paper, we propose a practice tracking-by-detection method composed by a convolutional neural network (CNN) detector and a multiple hypothesis tracking (MHT) tracker. In order to deal with complex scenes efficiently and accurately, it utilize modified pruning strategies and multiple data information, such as the kinematic and appearance features. And the experiments on the self-designed simulation radar datasets prove that this method has the ability to reach a good performance with both accuracy and a low computational cost.

Human activity classification is a critical part of intelligent human-computer interaction, which is promising in various applications. The radar system provides a complementary input when the visible light cannot work. It can effectively tackle darkness, occlusion, and non-line-of sight conditions. The frequency modulation induced by the micromotion of human movements can be captured and utilized by radar systems. Currently, the integration of deep learning methods and classical signal processing has been more and more prominent in the radar-based human behaviour sensing field. In this work, we propose a compact deep learning model to classify various human activities by the corresponding frequency modulation. Considering most deep neural networks are both data-hungry and resource-hungry, we integrate transfer learning and network pruning techniques to reduce the number of labelled training samples and computational burden. The experiments demonstrate that our method not only outperforms convolutional neural networks trained from scratch but also significantly slims its model size and computing operations.

In this paper, a deep learning framework for space-time adaptive processing is developed. Firstly, a set of clutter covariance matrixes (CCMs) are modeled based on the prior parameters of radar and navigation system with respect to all possible levels of non-ideal factors, and the columns of each CCM is formulated as undersampled noisy linear measurements of the sparse coefficients corresponding to angle-Doppler spectrum. Then the original spectrum coefficients, obtained by least-square estimation from the modeled CCMs and known space-time steering dictionary, are used as input to train the convolutional neural network (CNN). Meanwhile, the corresponding labels can be obtained by the exact spectrum of modeled CCMs via minimum variance distortionless response algorithm. Once trained, the CNN can be used to predict angle-Doppler spectrum coefficients that corresponds to a new measurement vector in near real time. Simulations results have demonstrated the superiority of the proposed method in both clutter suppression performance and computation efficiency.

Hidden Markov Model (HMM) is a commonly used modeling approach for the radar target recognition based on high-resolution range profile (HRRP). The HMM describes HRRP with the assumption of Gaussian observation probability, which results in the inaccurate description due to the non-Gaussian property of HRRP. To this end, we propose a hybrid HMM framework, which uses an extra model to fit the observation probability rather than making distribution assumptions. The extra model estimates the posterior probability of the state, and then can calculate the pseudo observation probability, which in turn modulates the other model parameters in HMM. The interplay between HMM and the extra model ultimately leads to a better description for observed data. In this paper, we use the support vector machine (SVM) and deep neural network (DNN) separately as the extra model and validate the superiority of our hybrid HMM in HRRP-based target recognition.

Deep learning has shown great success on object detection task for optical images, yet it’s rarely used in marine radar target detection. It is mainly due to the scarce of the labelled marine radar datasets for deep learning. To mitigate such effect, we build a pseudo-color marine radar dataset, which utilizes rich temporal information to remarkably enhance the accuracy of deep radar detectors. Meanwhile, the state-of-the-art one stage deep detector, YOLO family, is transferred into this field. Eventually, we modify the architecture of YOLOv3-tiny by attaching the prediction heads to shallower layers. Based on the pseudo-color dataset, our method, Shallow YOLOv3-tiny, is able to match the speed and model size of the standard YOLOv3-tiny while improving the accuracy significantly. Thus, our method is excellent at performing radar target detection tasks on embedded devices with low computing power and limited memory.

Nowadays, gesture recognition with radar is attracting wide attention from researchers and practitioners. The classification of an isolated and segmented gesture has been studied thoroughly. However, the detection, classification and segmentation of a series of gestures embedded in a data stream remains intractable. To address this problem, we develop a gesture recognition system based on millimetre-wave radar and deep learning. The radar measures range and Doppler features of gestures with high resolution. The data stream collected by radar is slightly pre-treated to suppress interference and extract information. A sliding window is used to slice those streams into appropriate data units, which are then fed to convolutional neural networks to estimate the probabilities of gesture types. By utilizing the change in those probabilities with time, the joint recognition and segmentation of gestures is realized. Experiments with real data shows that the recognition accuracy of 5 gestures is up to 92.48%.

The classification of intercepted radar signals has gained considerable attention in the field of electronic reconnaissance. Currently, the multi-function radar (MFR) is capable of transmitting complex and agile signals with different working modes, and the classification of radar waveforms using the traditional methods does not provide satisfactory results. Therefore, it is urgent to develop a new intelligent algorithm to recognize the working mode of MFR. In particular, this paper designs a novel feature extraction method to obtain the sequential relationship of the input signals. At the same time, a hierarchy of signal features is established to represent the signals layer by layer, and then the working mode of the emitter is determined effectively by unsupervised clustering method. The effectiveness and robustness of the proposed method is experimentally verified, through simulating the real complex electromagnetic environment and generating the signal samples of MFR.

Radar-based human activities recognition is still an open problem and is a key to detect anomalous behaviour for security and health applications. Deep learning networks such as convolutional neural networks (CNN) have been proposed for such tasks and showed better performance than traditional supervised learning paradigm. However, it is hard to deploy CNN networks to embedded systems due to the limited computational power available. From this point of concern, the use of a recurrent neural network (RNN) is proposed in this paper for human activities classification. We also propose an innovative data argumentation method to train the neural network using a limited number of data. The experiment shows that our network can achieve a mean accuracy of 94.3% in human activity classification.

Scattering of electromagnetic waves by discrete, randomly distributed objects inside a (finite or semi-infinite) slab is addressed. In general, the non-intersecting scattering objects can be of arbitrary form, material and shape with number density n0 (number of scatterers per volume). The main aim of this paper is to calculate the coherent reflection and transmission characteristics for this configuration. Typical applications of the results are found at a wide range of frequencies (radar up to optics), such as attenuation of electromagnetic propagation in rain, fog, and clouds etc. The integral representation constitutes the underlying framework of the solution of the deterministic problem, which then serves as the starting point for the solution of the stochastic problem. Conditional averaging and the employment of the Quasi Crystalline Approximation lead to a system of integral equations in the unknown expansion coefficients. The slab geometry implies a system of integral equations in the depth variable. Explicit solutions for tenuous media and low frequency approximations can be obtained for spherical obstacles. (Less)

This chapter provides an overview of the basic principles of ML, outlining the fundamental concepts that need to be applied correctly for a broad range of radar applications. We expect the reader to have background knowledge of basic linear algebra and probability theory, which form the foundations of ML. In Section 2.1, we describe the concept of learning from data and introduce the main categories of ML, namely, supervised and unsupervised learning. We also present different tasks that ML can tackle under each category and provide relevant radar-based examples. In Section 2.2, we briefly describe the various components of an ML algorithm. We present several fundamental techniques of supervised and unsupervised learning in Section 2.3. In Section 2.4, we define various performance assessment metrics and describe the design and evaluation of a learning algorithm. More recent learning approaches, such as variants of deep neural networks (DNNs), and more specific ML tools related to the various radar applications will follow in subsequent chapters of this book.

The automatic recognition of targets with radar is an ongoing research field for many years already. Since 2014, a new methodology based on deep neural networks is becoming more established within this field. This chapter gives a short overview with some examples of this short history of target recognition using deep learning (DL) and some comparative results with a basic implementation of a convolutional neural network (CNN). This network is applied to the commonly used Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset and to an inverse synthetic aperture radar (ISAR) dataset measured with the Tracking and Imaging Radar (TIRA) of Fraunhofer FHR.

In this chapter, two sparsity-driven algorithms of micro-Doppler analysis were presented for radar classification of rigid-body and nonrigid body targets, respectively. The first algorithm aimed to accurately estimate the micro-Doppler parameters of a rigid-body target. A parametric dictionary, which is dependent on the unknown angular speed of the target, was designed to decompose the radar echo into several dominant micro -Doppler components. By doing so, the problem of micro-Doppler parameter estimation was converted into the problem of sparse signal recovery with a parametric dictionary. To avoid the time-consuming full search, the POMP algorithm was presented by embedding the pruning process into the OMP procedure. Simulation results have demonstrated that the POMP algorithm can yield more accurate micro -Doppler parameter estimates and better time - frequency resolution in comparison with some well-recognized algorithms based on WVD and Hough transform. The second algorithm, referred to as the Gabor- Hausdorff algorithm, was presented for micro -Doppler feature extraction and applied to radar recognition of nonrigid body targets such as hand gestures. Taking advantage of the sparse properties of radar echoes reflected from dynamic hand gestures, the Gabor decomposition was used to extract the time -frequency locations and corresponding coefficients of the dominant signal components. The extracted micro-Doppler features were inputted into modified-Hausdorff-distance based NN classifier to determine the type of dynamic hand gestures. Experimental results based on real radar data have shown that the Gabor-Hausdorff algorithm outperforms the PCA-based and the DCNN-based methods in conditions of small training dataset.

An electromagnetic signal transmitted by a radar is reflected from a target then returns to the radar with the information of the target characteristics. Doppler information is commonly used to detect moving objects while suppressing clutter. In particular, the micro-Doppler signatures from nonrigid body motions contain diverse information regarding target movement [1-3]. Accordingly, the use of micro-Doppler signatures has a variety of defense, security, surveillance, and biomedicine applications, includ-ing airborne target classification, human detection, human activity classification, and drone detection.

Radar, short for "radio detection and ranging," was first conceived in the 1930s as means to provide early warning of approaching aircraft. It operates by transmitting an electromagnetic (EM) wave and processing the received echoes to measure the distance, velocity and scattering properties of objects and their surroundings. As radar transceivers themselves have become smaller, lighter in weight, and lower in cost due to advances in solid-state microelectronics, so has radar emerged as a key enabler in newer fields, such as automotive sensing and self-driving vehicle technology, gesture recognition for human-computer interfaces, and biomedical applications of remote health monitoring and vital sign detection. The proliferation of radar is indeed remarkable and presents a vast arena where deep learning, cybernetics, and artificial intelligence can enable disruptive technological advancements.

A variety of approaches for addressing the challenges involved in training DNNs for the classification of radar micro-Doppler signatures have been presented in this chapter. The performance metrics shown throughout the chapter reveal the impact of DNN training on the accuracy and target generalization performance of the network. Although high accuracies have been attained for the classification of as much as 12 different activity classes, open areas of research remain in regards to exploitation of radar datasets of opportunity and the generation of kinematically accurate, yet diverse, synthetic data. In this regard, adversarial learning provides opportunities, but ultimately both the training strategies and network architecture for training data synthesis must be designed uniquely for radar datasets. Training approaches must consider not only accurate modeling of target and clutter signatures but also must exploit constraints imposed by the physics of electromagnetic sensing to reduce complexity and increase performance. Advances in this area have the potential to greatly expand the application of RF sensors toward human motion recognition in both civilian and military applications.

A political and technical challenge for developing an operational ultrawideband (UWB) surveillance radar is the ability to comply with the many international regulations on the radio frequency (RF) spectrum. It is clear that the worldwide regulations on the RF spectrum have made it more difficult to get licenses for radar operation. More importantly, the recent auction of RF spectrum for civilian personal communications in several countries has reduced the traditional bands for radar spectrum access. This chapter will go over the beneficial uses of UWB surveillance radars, along with the attendant frequency allocation process. It is very difficult to get an RF allocation license for operating in any populated area. And the process is always changing, with the advent of personal communications, a widespread reliance on wideband communications, digital television and radio broadcast. A short history of the RF frequency allocation in recent time will show a summary of the process, and the story will present several lessons learned from obtaining licenses for UWB radars.

This chapter has provided an overview of the radar fundamentals, system design, prototyping, and signal processing of FMCW radars for micro-Doppler detection. The design principles of each part of an FMCW radar, including the radar transceiver, link budget, antenna design, etc., have been introduced and discussed. Besides the hardware design, range-Doppler and micro-Doppler effects in radar signal processing have also been discussed in detail. Finally, a brief introduction has been given in this chapter to cover the emerging applications for range-Doppler and micro-Doppler effects.

In this chapter, we have demonstrated the feasibility of sparsity -driven methods for micro -Doppler detection and classification. Two scenarios have been considered: the dictionary is pre -defined and fixed during the processing. Taking advantage of the sparse property of radar echo reflected from dynamic hand gestures, the Gabor dictionary and the OMP algorithm are used to extract the micro -Doppler features. The extracted features are inputted into modified-Hausdorff-distance-based nearest neighbour classifier to determine the type of dynamic hand gestures. Experimental results based on measured data show that the sparsity -driven method obtains recognition accuracy higher than 96% and outperforms the PCA-based and the DCNN-based methods in the conditions of small training dataset. The dictionary is learnt from the training data using dictionary learning methods. We used these sparse codes to detect multiple movers (humans moving along different trajectories) in the channel as well as to classify different types of human motions.

There has been a recent surge in activity related to bistatic synthetic aperture radar (SAR) systems. This is, in part, due to recent advances in hardware making such systems more attainable and the fact that the approach itself has the potential to provide the following unique properties: 1. Defeating counter-measures (such as jamming). 2. Improving detection performance by reduction of 'spiky' clutter. 3. Investigation of complex radar phenomenology (e.g. analysis of complex scattering environments, such as the inside of buildings). A SAR image is typically formed by having a moving radar system (e.g. on a spacecraft or an aeroplane) which transmits a series of pulses illuminating an area of the ground and records the energy reflected from the ground. It is then possible to combine these recorded reflected pulses into a map of radar cross section (RCS) as a function of 2D position within the illuminated area.

The micro-Doppler signature is a distinctive characteristic of micro-motion represented by intricate time-varying Doppler frequency modulations. It can be used to extract motion kinematic features for identifying an object of interest. In this Chapter, we introduce the micro-Doppler signature in radar and its applications for sensing micro-motion.

In this Chapter, we propose a novel dynamic monopulse radar (DMR) sensor with new localizing algorithm for both indoor positioning and surgical instrument positioning. The proposed DMR system consists of a steerable comparator, RSSI modules, and a microcontrol unit (MCU). Based on the engagement between radar techniques and phased array techniques, this proposed system can provide two-dimensional (2-D) angulation positioning using only a single positioning node.

This chapter reviews a recently developed technique for directly measuring the angular velocity of moving objects using a distributed radar aperture with interferometric processing. The use of a distributed interferometric radar provides a direct measurement method for the angular velocity, a quantity which can traditionally only be inferred over multiple measurements.

The use of radar technologies for biomedical research has potential benefits due to their effectiveness for noninvasively monitoring physiological responses and more humane treatments of living animals. It provides a useful tool for a more rapid, efficacious, and cost-effective extraction of animal's physiological data with great potential for clinical animal research, such as studying behavioral science (feeding rate or licking rate), monitoring the onset and progression of disease, and assessment of physiological responses to new treatments or drugs in a stress-free manner.

In this chapter, we discuss the advantages of exploiting photonic techniques in radar networks. More in detail, we first review the specific requirements of radar networks, focusing in particular on the coherence requested to the radar signals at the different network nodes. Then, we discuss the potentials of using photonics for distributing synchronization signals through optical fibers in distributed radar networks. Finally, we concentrate on the most performing radar network architecture, namely the centralized radar network, to report very recent experimental results enabled by photonics, and analyze the potential of a multi-input-multi-output (MIMO) processing approach in a radar network with widely separated antennas.

In this study, a new design of dual-polarized ultra-wideband (UWB) antenna is proposed. The antenna design contains a circular-ring slot radiator fed by two independently semi-arc-shaped microstrip feeding lines which can exhibit polarization diversity characteristic. A low-cost FR-4 dielectric (ε= 4.4, δ= 0.02) is used as the substrate. The characteristics of the dual-polarized UWB antenna are examined using both simulations and measurements and good results are achieved. An impedance bandwidth (S 11 ≤ −10 dB) of 2.5–10.2 GHz with 121% fractional bandwidth (FBW) is achieved for the design. However, for S 11 ≤ −6 dB, this value is more than 130% (2.2–11 GHz). The proposed UWB antenna offers good isolation, dual-polarized function, and sufficient efficiency which make it suitable for different applications such as radar and microwave imaging.

Fundamentally, this book seeks to capture the salient aspects of the exceedingly complex topic of radar and communication spectrum sharing, which is itself a menagerie of different problem spaces that depend on the particular goals one is attempting to achieve. The most prominent problem, and consequently the one most often first considered, is that of sharing between commercial communications and monolithic, stationary radars (e.g. for weather monitoring or air traffic control). Due to the sheer breadth of this topic, we do not attempt to codify all the many ways in which spectrum sharing could be performed. In fact, with the notion of radar spectrum sharing only rather recently becoming manifest due to the confluence of exponentially growing spectrum demand and emerging software-driven radio capabilities, one could well contend that we are now only glimpsing what will later be considered the earliest stages of spectrum sharing innovations.

The escalating interest on the topic of coexistence between radar systems and broadband communication devices is a direct consequence of the significant portion of the international radio spectrum currently allocated to radar systems. However, studies show that their spectrum occupancy is low in the spatial, temporal, and frequency domains. The most promising radar bands for shared use are the L, S, and C bands located in the 960-1400 MHz, 2.7-3.6 GHz, and 5.0-5.85 GHz frequency ranges, respectively. These frequencies are sufficiently low to avoid high power consumption and the usage of highly directional antennas, and sufficiently high to offer considerable bandwidth to commercial services. Furthermore, they are close to the cellular and ISM bands used for 2G/3G/4G and wireless local area networks (WLAN), respectively, facilitating the production of low-cost devices capable of using all these frequencies.

In this chapter, we analyze the general problem of a wideband radar that shares the same channel with a communication system, assuming that the communication band is divided into several frequency channels used for dynamic spectrum access (DSA). We specifically consider the case in which the radar is the secondary user (SU) and the communication system is the primary user (PU). In cognitive radio terminology, PUs are those who have higher priority or legacy rights to the usage of a specific part of the spectrum. SUs have lower priority and exploit the spectrum in such a way as to not cause interference to PUs. Therefore, SUs must possess some cognitive radio capabilities, such as an ability to sense the spectrum reliably to determine whether or not it is currently occupied by a PU. If the given channel becomes occupied by a PU, the SU must then move to an unused part of the spectrum, often referred to as white spaces or spectrum opportunities.

Analysis has shown that information exchange between radar and communications greatly enhances sharing performance. Once the current/future system state information is exchanged, radar and communication systems may use coordinated interference avoidance, mitigation, and amalgamation techniques that are otherwise impossible when the systems do not share information and are forced to reactively sense-andavoid in the spectrum. In addition, information exchange enables these systems to engage in joint co-designed and cooperating operating modes that open the door to amalgamated RC waveforms where the RF equipment can synergistically support each other's operation.

The objective of this chapter is to provide a high-level perspective on the growing conflict over use of the radio-frequency (RF) spectrum, a precious and highly sought resource extending from below 1 MHz to above 100 GHz, caused by the accelerating demand for consumer use via 4G and soon-to-be 5G wireless communications. The world at large now faces serious spectrum-compatibility problems that require new and innovative solutions-increased spectral congestion and crowding are especially challenging. However, anticipated improvements in electromagnetic (EM) systems up to 300 GHz are beginning to be realized. Less restrictive constraints on communication systems, inherent in one-way propagation paths and much less expensive components, have allowed that community to design and develop more diverse waveforms and systems. Consequently, commercial cellular systems are proliferating at incredible rates, resulting in extremely spectrally dense environments and fierce competition for spectrum that traditionally has been the almost exclusive province of radars as primary legal users. For radar applications, however, the promise is being realized much more slowly, and the inundation of communication devices from the commercial sector has caused significant radar-communication interference problems. In addition, radar and communication systems are important components of military operations, and advances in waveform-diversity signal and data-processing techniques that are likewise relevant to spectrum sharing offer the promise of significantly improved performance.

This chapter has provided an introduction to the spectrum crunch problem from a radar perspective. At one level it can be said that the problem is already severe. There is ever-greater competition for a resource that is strictly finite, and radar is only one voice among many pressing their case. All users have a need for greater bandwidth, and the only thing that can be said with certainty is that the problem is only going to get worse.Yet if spectrum usage were measured at a given point as a function of frequency, time, space and polarisation, it would certainly be found that the spectrum is currently not being used that efficiently. There is therefore great potential for approaches aimed at using the spectrum in a more efficient and dynamically controlled manner. The regulatory framework has thus far taken a relatively conservative approach. However, it is important to have a proper quantitative understanding of the effect of interference of one service upon another in order to adopt appropriate regulation measures, rather than taking the view that no service should ever occupy the same part of the spectrum as any other. To date, a number of novel radar technology approaches have been introduced, including improvements to the spectral purity of transmitters, intelligent, cognitive approaches to dynamic frequency allocation, passive sensing based on the emissions of other RF applications, and even through learning to mimic the behaviour of echolocating animals. These topics and others are developed in the chapters that follow, and give some cause for optimism that a solution can be found.

This chapter provides an overview of recent advances in radar-embedded communication signals. Several information-embedding techniques have recently been successful in establishing dual-function systems that simultaneously perform both radar and communication functions. Similarto single-function communication platforms, dual function systems should provide secure communications to protect user signals from being intercepted by other users or unintended receivers. Information embedding into the emission of multiple-input (MIMO) multiple-output radar provides a means for broadcast communications to be secondary to the primary radar function of the dual system.