Micro-Doppler Radar and its Applications (Radar, Sonar and Navigation)

Cover
Contents
About the editors
List of reviewers
Preface
1 Multistatic radar micro-Doppler
	1.1 Introduction
	1.2 Bistatic and multistatic radar properties
	1.3 Description of the radars and experimental setups
	1.4 Analysis of multistatic radar data
		1.4.1 Micro-Doppler signatures
		1.4.2 Feature diversity for multistatic radar systems
		1.4.3 Multiple personnel for unarmed/armed classification
		1.4.4 Simple neural networks for multistatic radar data
		1.4.5 Personnel recognition using multistatic radar data
		1.4.6 Multistatic drone micro-Doppler classification
	1.5 Concluding remarks
	Acknowledgements
	References
2 Passive radar approaches for healthcare
	2.1 Introduction
		2.1.1 Ambient assisted living context
		2.1.2 Sensor technologies in healthcare
	2.2 Passive radar technology for human activity monitoring
		2.2.1 Current radar technology research with application to healthcare
		2.2.2 Passive radar technologies for activity monitoring
	2.3 Human activity recognition using Wi-Fi passive radars
		2.3.1 Fundamentals of Wi-Fi passive radars
		2.3.2 Micro-Doppler activity recognition
			2.3.2.1 Signal model
			2.3.2.2 Micro-Doppler feature extraction
			2.3.2.3 Classification methods
			2.3.2.4 Experiments and results
			2.3.2.5 Summary
		2.3.3 UsingWi-Fi CSI informatics to monitor human activities
			2.3.3.1 From micro-Doppler (to channel state information (CSI)
			2.3.3.2 Wi-Fi CSI signal capture and parameter extraction
			2.3.3.1 From micro-Doppler (μD) to channel state information (CSI)
			2.3.3.3 Wi-Fi CSI-enabled healthcare applications
	2.4 The future for passive radar in healthcare
	2.5 Conclusions
	References
3 Sparsity-driven methods for micro-Doppler detection and classification
	3.1 Introduction
	3.2 Fundamentals of sparse signal recovery
		3.2.1 Signal model
		3.2.2 Typical algorithms for sparse signal recovery
			3.2.2.1 Convex optimisation
			3.2.2.2 Bayesian sparse signal recovery
			3.2.2.3 Greedy algorithms
		3.2.3 Dictionary learning
	3.3 Sparsity-driven micro-Doppler feature extraction for dynamic hand gesture recognition
		3.3.1 Measurement data collection
		3.3.2 Sparsity-driven dynamic hand gesture recognition
			3.3.2.1 Extracting time–frequency trajectory
			3.3.2.2 Clustering for central time–frequency trajectory
			3.3.2.3 Nearest neighbour classifier based on modified Hausdorff distance
		3.3.3 Experimental results
			3.3.3.1 The recognition accuracy versus the sparsity
			3.3.3.2 The recognition accuracy versus the size of training dataset
			3.3.3.3 Recognition accuracy for unknown personnel targets
			3.3.3.4 Computational costs considerations
	3.4 Sparsity-based dictionary learning of human micro-Dopplers
		3.4.1 Measurement data collection
		3.4.2 Single channel source separation of micro-Dopplers from multiple movers
		3.4.3 Target classification based on dictionaries
	3.5 Conclusions
	References
4 Deep neural networks for radar micro-Doppler signature classification
	4.1 Radar signal model and preprocessing
		4.1.1 2D Input representations
		4.1.2 3D Input representations
		4.1.3 Data preprocessing
			4.1.3.1 Dimensionality reduction
			4.1.3.2 Mitigation of clutter, interference, and noise
	4.2 Classification with data-driven learning
		4.2.1 Principal component analysis
		4.2.2 Genetic algorithm-optimised frequency-warped cepstral coefficients
		4.2.3 Autoencoders
		4.2.4 Convolutional neural networks
		4.2.5 Convolutional autoencoders
	4.3 The radar challenge: training with few data samples
		4.3.1 Transfer learning from optical imagery
		4.3.2 Synthetic signature generation from MOCAP data
		4.3.3 Synthetic signature generation using adversarial learning
			4.3.3.1 Case study: auxiliary conditional GANs
			4.3.3.2 Kinematic fidelity ofACGAN-synthesised signatures
			4.3.3.3 Diversity ofACGAN-synthesised signatures
			4.3.3.4 PCA-based kinematic sifting algorithm
	4.4 Performance comparison of DNN architectures
	4.5 RNNs and sequential classification
	References
5 Classification of personnel for ground-based surveillance
	5.1 Introduction
	5.2 Modelling and measuring human motion
		5.2.1 Human gait
			5.2.1.1 Human walking model
			5.2.1.2 Human running model
			5.2.1.3 Other gait models
		5.2.2 Measuring the signature of human gait
			5.2.2.1 Micro-Doppler signature
		5.2.3 Visualisation of the micro-Doppler signature
		5.2.4 Motion capturing activities on human motion
	5.3 Model-driven classification
		5.3.1 Introduction to model-driven classification methods
		5.3.2 Results of model-based classification
		5.3.3 Particle-filter-based techniques for human gait classification and parameter estimation
			5.3.3.1 Particle filter
			5.3.3.2 Human gait implementation
			5.3.3.3 Classification and parameters estimation results
			5.3.3.4 Conclusion
	5.4 Data-driven classification
		5.4.1 Deep supervised learning for human gait classification
			5.4.1.1 Convolutional neural networks
			5.4.1.2 Recurrent neural networks
		5.4.2 Deep unsupervised learning for human gait classification
			5.4.2.1 Generative Adversarial Networks
			5.4.2.2 AdversarialAutoencoders
		5.4.3 Conclusion
	5.5 Discussion
	References
6 Multimodal sensing for assisted living using radar
	6.1 Sensing for assisted living: fundamentals
		6.1.1 Radar signal processing: spectrograms
		6.1.2 Wearable sensors: basic information
	6.2 Feature extraction for individual sensors
		6.2.1 Features from radar data
			6.2.1.1 Automatic feature extraction for radar
			6.2.1.2 Handcrafted features
		6.2.2 Features from wearable sensors
		6.2.3 Classifiers
	6.3 Multi-sensor fusion
		6.3.1 Principles and approaches of sensor fusion
		6.3.2 Feature selection
	6.4 Multimodal activity classification for assisted living: some experimental results
		6.4.1 Experimental setup
		6.4.2 Heterogeneous sensor fusion: classification results for multimodal sensing approach
			6.4.2.1 Magnetic sensor and radar results
			6.4.2.2 IMU and radar results
	6.5 More cases of multimodal information fusion
		6.5.1 Sensor fusion with same sensor: multiple radar sensors
	6.6 Conclusions
	References
7 Micro-Doppler analysis of ballistic targets
	7.1 Introduction
	7.2 Radar return model at radio frequency
		7.2.1 Cone
			7.2.1.1 Cone plus fins
		7.2.2 Cylinder
		7.2.3 Sphere
	7.3 Laboratory experiment
	7.4 Classification framework
		7.4.1 Feature vector extraction
			7.4.1.1 ACVD-based feature vector approach
			7.4.1.2 2D Signature-based feature vectors
		7.4.2 Classifier
	7.5 Performance analysis
		7.5.1 ACVD approach
		7.5.2 pZ Moments approach
		7.5.3 Kr moments approach
		7.5.4 2D Gabor filter approach
		7.5.5 Performance in the presence of the booster
		7.5.6 Average running time
	7.6 Summary
	References
8 Signatures of small drones and birds as emerging targets
	8.1 Introduction
		8.1.1 Classes and configuration of UAVs
		8.1.2 Literature review
	8.2 Electromagnetic predictions of birds and UAVs
		8.2.1 Electromagnetic properties, size and shape
		8.2.2 RCS predictions
	8.3 Target analysis using radar measurements
		8.3.1 Body RCS
		8.3.2 Rotor RCS
		8.3.3 Maximum CPI
		8.3.4 Micro-Doppler analysis
		8.3.5 Micro-Doppler analysis on polarimetric data
	8.4 Radar system considerations
		8.4.1 Carrier frequency and polarisation
		8.4.2 Bandwidth
		8.4.3 Waveform and coherency
		8.4.4 Pulse repetition frequency
		8.4.5 Pencil-beam or ubiquitous radar
	8.5 Classification methods
	8.6 Conclusion
	Acronyms
	References
9 Hardware development and applications of portable FMCW radars
	9.1 FMCW radar fundamentals
	9.2 Radar transceiver
		9.2.1 Transmitter
		9.2.2 Receiver
	9.3 Antenna and antenna array
		9.3.1 Beamforming
		9.3.2 Two-way pattern and MIMO
	9.4 Radar link budget analysis
	9.5 FMCW radar signal processing
		9.5.1 Range processing
		9.5.2 Range-Doppler processing
		9.5.3 Micro-Doppler
	9.6 Applications of micro-Doppler effects
		9.6.1 Gesture recognition
		9.6.2 Fall detection
		9.6.3 Human activity categorising
	9.7 Summary
	References
10 Digital-IF CW Doppler radar and its contactless healthcare sensing
	10.1 Principles of digital-IF Doppler radar
	10.2 Overview of RF layer
	10.3 Implementation of digital-IF layer
		10.3.1 Direct IF sampling
		10.3.2 Digital quadrature demodulation
		10.3.3 Decimation
	10.4 DC offset calibration technique
	10.5 Applications to healthcare sensing
		10.5.1 Contactless beat-to-beat BP estimation using Doppler radar
		10.5.2 Multi-sensor-based sleep-stage classification
	10.6 Summary
	References
11 L1-norm principal component and discriminant analyses of micro-Doppler signatures for indoor human activity recognition
	11.1 Introduction
	11.2 Radar signal model
	11.3 L1-PCA-based classification
		11.3.1 L1-norm PCA
		11.3.2 L1-PCA through bit flipping
		11.3.3 Classifier
		11.3.4 Illustrative example
	11.4 L1-LDA-based activity classification
		11.4.1 Problem formulation
		11.4.2 L1-LDA algorithm
		11.4.3 Classifier
		11.4.4 Illustrative example
	11.5 Discussion
	11.6 Conclusion
	References
12 Micro-Doppler signature extraction and analysis for automotive application
	12.1 Introduction and state of the art
	12.2 Micro-Doppler analysis in automotive radar
		12.2.1 Target detection techniques
		12.2.2 Tracking techniques
		12.2.3 Track-based spectrogram extraction
		12.2.4 Spectrogram processing
		12.2.5 Feature extraction and classification
	12.3 Experimental validation
		12.3.1 Experimental radar system
		12.3.2 Experimental set-up
		12.3.3 Processing
			12.3.3.1 Classification results
	12.4 Practicality discussion
	12.5 Conclusion
	References
Conclusion
Index
Back Cover