Micro-Doppler Parameter Estimation Based On Parameterized Sparse Representation And Its Application

When the radar detects a moving target in a radar scenario,there is a frequency offset between the carrier frequencies of received signal and transmitted signal.This phenomenon is known as the Doppler effect.In addition,the micro motions of the target or its parts will cause additional frequency modulation on the received signal.This phenomenon is called as the micro-Doppler effect.Micro-Doppler signals generated by the micro-Doppler effect contain much information about characteristic parameters of the target,such as electromagnetic scattering characteristic,geometry and motion state.If these characteristic parameters can be estimated correctly,they will benefit target detection and recognition.However,the micro-Doppler signal is always mixed with the Doppler signal.We firstly need to recover the micro-Doppler signal from the mixed signal and then estimate the parameters of micro-Doppler signal.Therefore,according to the difference between statistical properties of the micro-Doppler signal and the Doppler signal in time-frequency domain,this thesis firstly studies a Doppler signal reconstruction technique based on compressive sensing in order to recover the micro-Doppler signal;Thereafter,to improve the performance of micro-Doppler signal parameter estimation,this thesis addresses a micro-Doppler parameter estimation algorithm based on parameterized sparse representation,and applies it into the field of helicopter model recognition.The main works of this thesis are given as follows:1.Study of the micro-Doppler signal recovery method based on the statistical property in time-frequency domain.Firstly,we use the histogram statistics to extract the time-frequency coefficients in time-frequency domain which contains only the information of Doppler signal,because the micro-Doppler signal and the Doppler signal have different distribution characteristics.Then we reconstruct the Doppler signal based on compressed sensing theory.Finally,we recover the micro-Doppler signal according to the reconstructed Doppler signal and the original mixed signal,.2.Study of the micro-Doppler parameter estimation algorithm(IPOMP algorithm)based on the parameterized sparse representation of micro-Doppler signal.The micro-Doppler signal contains three main parameters:angular frequency and Doppler amplitude and initial phase.With the sparse representation,the micro-Doppler parameter estimation problem is translated into a sparse signal recovery problem.As the angular frequency has the most influence on the characteristics of micro-Doppler signal,we parameterize the angular frequency in sparse representation.On the other hand,to reduce the computational cost,wediscretize the Doppler amplitude and initial phase in sparse representation.For the parameterization of angular frequency,the proposed algorithm firstly discretizes an interval of angular frequency where the micro-Doppler signal may exist,and finds out the discrete grid of angular frequency which is closest to the real angular frequency via matching pursuit algorithm;Then the real micro-Doppler signal is approximated by the first-order Taylor series at the closest discrete grid,and the angular frequency estimation problem is translated into the estimation problem of the deviation between the real angular frequency and the closest discrete grid.By means of the alternate optimization startegy,the proposed algorithm can effectively estimate the closest discrete grid and the deviation.3.Application of the IPOMP algorithm into the field of helicopter model recognition.This thesis firstly analyzes the model of micro-Doppler signals generated by the helicopter's main rotor;and then exploits the IPOMP algorithm to estimate the angular frequency of the micro-Doppler signal;and finally utilizes the angular frequency estimate to recognize the helicopter model.The simulation results show that the proposed algorithm has high model recognition rate for the helicopters with the same number of blades and the close rotation rate.