WIXLIB

2015 The MathWorks, Inc.1

Machine Learning for Radar & EW서기환 과장 2015 The MathWorks, Inc.2

Agenda Radar and EW modeling Synthesizing data for Machine Learning workflows Machine Learning ExamplesAntenna/Array/RFSignalProcessingTracking andSensor FusionTargetClassificationSchedulingand Control3

Radar and EW Simulation and Modeling ArchitectureWaveformGeneratorλRadarScheduler andTracker TransmitArrayGt𝑃𝑡 𝐺𝑡 𝐺𝑟 𝜆2 𝜎𝑃𝑟 4𝜋 3 𝑅𝑡2 𝑅𝑟2 𝐿SignalProcessing TransmitterPtReceiverGrEnvironment,Targets, &Interference (L, 𝝈, 𝑹𝒓 , 𝑹𝒕)ReceiveArrayGrFunctions for calculations and analysisApps for common workflowsParameterized components for system modelingEasy path to increased fidelity for antenna and RF designCode generation for deployment4

Design Phased Array AntennasDesign an arrayTaper/thin arraysDesign subarraysModel imperfectionsImport antenna patternsSynthesize arraysModel failuresModel mutual coupling5

Design Radar and EW SystemsData cube processingSpatial signal processingDetectionsTargets & EnvironmentPolarizationWidebandScenario visualizationCode generationHDL6

Agenda Radar and EW modeling Synthesizing data for Machine Learning workflows Machine Learning ExamplesAntenna/Array/RFSignalProcessingTracking andSensor FusionTargetClassificationSchedulingand Control7

Machine Learning WorkflowAccess and ExploreDataPreprocess DataDevelop PredictiveModelsIntegrate Analytics withSystemsFilesWorking withMessy DataModel Creation e.g.Machine LearningDesktop AppsDatabasesData prise nEmbedded Devicesand Hardware8

Synthesize Radar Data for Machine LearningCollect dataTrain on dataMeasure accuracySynthesize dataCreate data setLearnValidate9

Synthesize Received Radar Signals10

Define a Backscatter Target with Angle and Frequencyrcs cyl cylinderrcs(r1,r2,H,c,fc,az,el);11

Model Basic Shapes12

Model Extended Targets with Multiple Scatters13

Agenda Radar and EW modeling Synthesizing data for Machine Learning workflows Machine Learning ExamplesAntenna/Array/RFSignalProcessingTracking andSensor FusionTargetClassificationSchedulingand Control14

Machine Learning for Radar ExamplesSynthesize returns (radar cross section)Synthesize micro-Doppler ynthesize waveformsClassificationStatistics and Machine Learning Toolbox15

Signal Processing and Wavelets for Feature Extraction Signal Manipulation– Signal Analysis App Time-Frequency Analysis Capabilities– Short Time Fourier Transform– Continuous Wavelet Transform– Synchrosqueezing Multiresolution Analysis Capabilities– Discrete Wavelet Analysis– Wavelet Packets16

Signal Analyzer AppAnalyze signals in time, frequency and time-frequency domains With the Signal Analyzer you can :– Import multichannel signals– Explore signals jointly in time-frequency domain– Zoom and pan signals17

Identifying Features in Real World SignalsQuantify time-varying signals in frequency domain Characterizing signal features in spectral domain is oftenchallenging as one needs to appropriate tools Accurate time-frequency measurements are possibleusing wavelet based time frequency analysis techniques Features once identified, can be extracted from signalsfor further processing In this demo, we will characterize features in EKG signalsusing Continuous Wavelet Transform» Demo18

Filtering Frequency Localized Components Sometimes unwanted signals can get capturedduring signal acquisition process A traditional filter cannot be used if the frequencyrange of the interference lies within the frequencyrange of the signal Unwanted components can be localized jointly intime and frequency using wavelets and removed19

Wavelet Synchrosqueezing For certain non-stationary signals, wavelet Synchrosqueezing can beused to identify and extract signal modes Wavelet Synchrosqueezing helps extract signal components fromlocalized regions of time frequency plane20

Example 1: Radar Echoes from Cylinder and Cone21

Generate law radar data from modelsTrain ondataSynthesizedataCreate data setLearnRandomize parametersGenerate many data sets22

Feature ExtractionExtract features: 700 samples/object - 8 samples/object with Wavelet Transform23

Testing Against Training Data24

Classification Learner AppAlgorithms for trainingMATLAB code genClassification metrics25

Example 2: Micro-Doppler for DronesMicro-Doppler returnsRange-Doppler from Parrot QuadcopterWe can identify: Rotation rate Number of blades Tip velocity Blade length26

Synthesize Micro-Doppler MotionRadar returnRange-Doppler of bladeMicro-Doppler Time-frequencyRotation rateNumber of bladesTip velocityBlade length27

Micro-Doppler in Time-Frequency DomainDetection of small UAV helicopters using micro-DopplerDavid Tahmoush28

Example 3: Waveform Modulation ID for RWR29

Radar & EW Classification – WorkflowAccess and ExploreDataFilesDatabasesPreprocess DataWorking&withStatisticsMachineMessy DataLearning ToolboxWaveletToolboxData Reduction/TransformationPhased Array SystemToolboxSensorsSignal ProcessingToolboxFeatureExtractionDevelop PredictiveModelsIntegrate Analytics withSystemsModel Creation e.g.Machine LearningDesktop AppsParameterOptimizationEnterprise ScaleSystemsStatistics & MachineLearning ToolboxModelValidationMATLAB CoderEmbedded Devicesand HardwareMATLAB30

Key Takeaways Radar and EW modeling Synthesizing data for Machine Learning workflows Machine Learning Examples– Synthesize Return, Micro-Doppler, Waveform– Classification: Target, Radar31

Resources to Help You Get Started매트랩과 머신러닝 (eBook)32

Machine Learning for Radar Examples Synthesize returns (radar cross section) Synthesize micro-Doppler (Time-frequency) Statistics and Machine Learning Toolbox Signals Features Time-frequency Etc. Classification Synthesize waveforms