کاربردهای یادگیری عمیق در رادارهای کوتاه‌برد

This exciting new resource presents emerging applications of artificial intelligence and deep learning in short-range radar. The book covers applications ranging from industrial, consumer space to emerging automotive applications. The book presents several human-machine interface (HMI) applications, such as gesture recognition and sensing, human activity classification, air-writing, material classification, vital sensing, people sensing, people counting, people localization and in-cabin automotive occupancy and smart trunk opening. The underpinnings of deep learning are explored, outlining the history of neural networks and the optimization algorithms to train them. Modern deep convolutional neural network (DCNN), popular DCNN architectures for computer vision and their features are also introduced. The book presents other deep learning architectures, such as long-short term memory (LSTM), auto-encoders, variational auto-encoders (VAE), and generative adversarial networks (GAN). The application of human activity recognition as well as the application of air-writing using a network of short-range radars are outlined. This book demonstrates and highlights how deep learning is enabling several advanced industrial, consumer and in-cabin applications of short-range radars, which weren't otherwise possible. It illustrates various advanced applications, their respective challenges, and how they are been addressed using different deep learning architectures and algorithms. Artech House Radar Series Deep Learning Applications of Short-Range Radars 2 Contents 6 Preface 14 1 Introduction to Radar Signal Processing 18 1.1 Types of Radar 20 1.1.1 CW Radar 20 1.1.2 Modulated CW Radar 20 1.1.3 Impulse UWB Radar 24 1.1.4 Other Short Range Radars 24 1.2 Waveform Design and Ambiguity Function 26 1.3 System Concept 30 1.4 Target Model 33 1.5 3D Data-cube Processing 37 1.5.1 1D Processing 39 1.5.2 2D Range Doppler Images 42 1.5.3 Range Cross-Range Images 45 1.6 Detection Strategy and Clustering 46 1.6.1 Detection Algorithm 46 1.6.2 Clustering 49 1.7 Parameter Estimation and Cramer-Rao Bound 53 1.8 Tracking 55 1.8.1 Track Management 55 1.8.2 Track Filtering 56 1.9 Applications of Short-Range Radar 65 1.10 Problems 70 References 71 2 Introduction to Deep Learning 76 2.1 Perceptron 77 2.2 Multilayer Perceptron 82 2.2.1 Training 84 2.2.2 Activation Functions 87 2.2.3 Optimizers 90 2.2.4 Types of Models 92 2.3 Convolutional Neural Networks 92 2.3.1 Convolution Layer 92 2.3.2 Popular Architectures 95 2.3.3 Transfer Learning 103 2.4 LSTM 104 2.5 Autoencoders 107 2.6 Variational Autoencoder 109 2.7 Generative Adversarial Network 112 2.8 Robust Deep Learning 117 2.9 Problems 118 References 119 3 Gesture Sensing and Recognition 122 3.1 Introduction 122 3.1.1 RelatedWork 124 3.2 Gesture Sensing/Detection 125 3.3 Micro-Gestures 128 3.3.1 System Parameters 128 3.3.2 Micro-Gesture Set 129 3.4 2D All CNN-LSTM 129 3.4.1 Architecture and Learning 130 3.4.2 System Evaluation 133 3.5 3D CNN and Pseudo-3D CNN 134 3.5.1 3D CNN Architecture and Learning 135 3.5.2 Pseudo-3D CNN Architecture and Learning 135 3.6 Meta-Learning 136 3.6.1 Architecture and Learning 138 3.6.2 System Evaluation 141 3.7 Macro-Gestures 148 3.7.1 System Parameters 148 3.7.2 Macro-Gesture Set 148 3.8 Unguided Attention 2D DCNN-LSTM 150 3.8.1 Architecture and Learning 150 3.8.2 System Evaluation 152 3.9 FutureWork and Direction 154 3.10 Problems 155 References 155 4 Human Activity Recognition and Elderly-Fall Detection 158 4.1 Introduction 158 4.1.1 RelatedWork 159 4.2 Preprocessing for Feature Image 160 4.2.1 Fast-Time FFT 160 4.2.2 Coherent Pulse Integration 161 4.2.3 MTI Filtering 161 4.2.4 Adaptive Detection Thresholding 162 4.2.5 Euclidean Clustering 162 4.2.6 Kalman Filter 162 4.3 Input Feature Images 163 4.3.1 Range Spectrogram 164 4.3.2 Doppler Spectrogram 164 4.3.3 Video of RDI 166 4.4 Human Activity Data Set 166 4.5 DCNN Activity Classification 167 4.5.1 Architecture and Learning 167 4.5.2 Results and Discussion 170 4.6 Bayesian Classification 172 4.6.1 Integrated Classifier and Tracker 175 4.6.2 Results and Discussion 179 4.7 Fall-Motion Recognition 183 4.7.1 Architecture and Learning 185 4.7.2 Deformable CNN 186 4.7.3 Loss Function 188 4.7.4 Results and Discussion 190 4.8 FutureWork and Directions 191 4.9 Problems 193 References 193 5 Air-Writing 196 5.1 Introduction 196 5.2 Radar Network Placement 198 5.3 Preprocessing 200 5.3.1 Coherent Pulse Integration 200 5.3.2 Moving Target Indication Filtering 200 5.3.3 Target Detection and Selection 201 5.3.4 Localization with Trilateration 202 5.3.5 Trajectory Smoothening Filters 204 5.4 Setup and Characters 205 5.4.1 Character Set 205 5.4.2 System Parameters 205 5.4.3 Setup and Data Acquistion 205 5.5 LSTM 207 5.5.1 Architecture 208 5.5.2 Loss Function: CTC 211 5.5.3 Design Considerations 213 5.5.4 Performance Evaluation 213 5.6 Deep Convolutional Neural Networks 214 5.6.1 Architecture 214 5.6.2 Weight Initialization 215 5.6.3 Learning Schedule 215 5.6.4 Data Augmentation 215 5.6.5 Performance Evaluation 216 5.7 1D CNN-LSTM 217 5.7.1 Architecture 217 5.7.2 Performance Evaluation 217 5.8 FutureWork and Directions 218 5.9 Problems 220 References 220 6 Material Classification 222 6.1 Introduction 222 6.1.1 RelatedWork 223 6.2 Features: Range Angle Images 224 6.3 Deep Convolutional Neural Networks 227 6.3.1 Architecture and Learning 229 6.3.2 Design Considerations 231 6.3.3 Results and Discussion 232 6.4 Siamese Network 234 6.4.1 Architecture and Learning 235 6.4.2 Design Considerations 236 6.4.3 Results and Discussion 237 6.5 FutureWork and Directions 239 6.6 Problems 239 References 240 7 Vital Sensing and Classification 242 7.1 Introduction 242 7.2 Vital Signal Fundamentals 244 7.2.1 Preprocessing Steps 247 7.3 Heart Rate Estimation through a Deep-Learning Approach 251 7.3.1 GAN-Based Data Augmentation 253 7.3.2 Results and Discussions 255 7.4 Adaptive Signal Processing with a Tracking Approach 257 7.4.1 Algorithm 258 7.4.2 Results and Discussion 261 7.5 IQ Signal Evaluation using Deep Learning 264 7.5.1 Deep Learning Architecture 266 7.5.2 Results and Discussion 266 7.6 FutureWork and Direction 267 7.7 Problems 268 References 268 8 People Sensing, Counting,and Localization 272 8.1 Introduction 272 8.2 Presence Sensing: Signal Processing Approach 277 8.2.1 Challenges 277 8.2.2 Solution 278 8.3 Presence Sensing: Deep Learning Approach 280 8.3.1 Challenges 280 8.3.2 Solution 282 8.3.3 Results and Discussion 285 8.4 People Counting: Signal Processing Approach 286 8.4.1 Challenges 286 8.4.2 Solution 288 8.5 People Counting: Deep Learning Approach 293 8.5.1 Data Preparation and Processing 296 8.5.2 Solution: Framework and Learning 298 8.5.3 Results and Discussion 302 8.6 People Detection and Localization: Signal Processing Approach 303 8.7 People Detection and Localization: Deep Learning Approach 307 8.7.1 Challenges 307 8.7.2 Architecture and Learning 309 8.7.3 Results and Discussion 312 8.8 FutureWork and Direction 318 8.9 Problems 319 References 319 9 Automotive In-Cabin Sensing 324 9.1 Introduction 324 9.2 Smart Trunk Opening 325 9.2.1 Challenges 325 9.2.2 Solution 326 9.2.3 Results and Discussion 331 9.3 Vehicle Occupancy Sensing 331 9.3.1 Challenges 331 9.3.2 Solution 333 9.4 Federated Learning 336 9.4.1 Challenges 336 9.4.2 Solution 338 9.5 FutureWork and Direction 339 9.6 Problems 339 References 340 About the Authors 342 Index 344 Radar;,Deep,learning;,Short-range,radar;,978-1-63081-746-6;,Artech,House Radar,Deep learning,Short-range radar,978-1-63081-746-6,Artech House This new resource covers various emerging applications of short range radars, including people counting and tracking, gesture sensing, human activity recognition, air-drawing, material classification, object classification, vital sensing by extracting features such as range-Doppler Images (RDI), range-cross range images, Doppler Spectrogram or directly feeding raw ADC data to the classifiers. The book also presents how deep learning architectures are replacing conventional radar signal processing pipelines enabling new applications and results. It describes how deep convolutional neural networks (DCNN), long-short term memory (LSTM), feedforward networks, regularization, optimization algorithms, connectionist temporal classification (CTC) are enabling these applications