Time-frequency Analysis And Parameter Estimation Of Nonlinear Frequency Modulation Signals

In modern signal processing,nonlinear frequency modulation(NLFM)signal is a typical non-stationary signal,it has a wide range of applications in many fields,such as communication,sonar,radar,biomedical science,seismic signal processing and other fields.Therefore,it is of great significance to study the time-frequency analysis and parameter estimation of NLFM signal.At present,domestic and foreign scholars have perfect theory for LFM signal analysis,and the research of NLFM signal is still in the initial stage,it is one of the research hotspots in signal processing.In this thesis,the time-frequency analysis and parameter estimation problems of NLFM signal are studied,the research subjects including sinusoidal frequency modulation(SFM)signal,high-oder polynomial phase signal(PPS).The specific work contents are done as follows:(1)Aiming at the cross-term interference problem of SFM signal Wigner-Ville Distribution(WVD),a new time-frequency analysis method based on Gabor atom decomposition is studied.This method makes full use of the high time-frequency aggregation of Gabor atoms,then the signal decomposition is realized by matching pursuit(MP)algorithm,combined with the improved genetic algorithm to search the best atoms,in the process of signal decomposition,and the cross-term interference is effectively restrained by a series of the best matching atoms.(2)In order to solve the parameter estimation problem of SFM signal in a low signal-to-noise ratio condition,a parameter method based on cyclic autocorrelation function(CAF)algorithm is studied.This method utilizes the cyclic autocorrelation function of SFM signal to periodically shows peak characteristics when the cyclic frequency is zero,the modulation frequency,carrier frequency and modulation index are estimated by the refined peak search and delay multiplication.(3)For the self-intersection and intersection interference problems of multicomponent PPS,a new time-frequency analysis method based on multi-scale chirplet decomposition is studied.Firstly,this method projects signals onto the parameterized chirplet base function for signal decoposition,searching for the best base function by the delay correlation demodulation method of FRFT,and then through the Wigner-Ville transform and the optimal path connection pursuit algorithm,the best functions constitutes the largest energy signals component and power distribution in returns.(4)In order to solve the parameter problem of high order PPS,a new parameter estimation method based on time-frequency curve fitting is studied.This method utilizes the smooth pseudo Wigner-Ville transform to suppress the signal cross-term interference,and extracts the signal instantaneous frequency curve from the maximum of time-frequency distribution,then the signal parameter estimation is realized by the least square and curve fitting algorithm.