Time-frequency Analysis Of Non-stationary Signal And Its Application In Frequency Hopping Signal Estimation

In modern electronic warfare,the accuracy and comprehensiveness of information acquisition play a decisive role in the final result.As carriers of information,the signals in engineering applications are almost non-stationary signals.In order to analyze the signals facilitatly,non-stationary signals are often regarded as the linear combination of multiple component signals.As an effective method for processing non-stationary signals,timefrequency analysis has been widely used.And the enhancement of time-frequency aggregation is very important for signal analysis and parameters estimation.Based on the short-time Fourier transform(STFT)and synchrosqueezing transform(SST),this thesis focuses on the adaptive signal processing and instantaneous frequency(IF)estimation,to provide stronger support for the signal recognition and parameters estimation.Firstly,the thesis introduces the theories of STFT-based SST(FSST)and second-order FSST(FSST2)for overcoming the shortcomings of traditional time-frequency analysis,such as low time-frequency aggregation,cross terms,and irreversibility.To achieve the goal of multiresolution analysis,the thesis defines the adaptive STFT with time-varying parameters,and derives the adaptive reference frequency to propose the adaptive FSST algorithm.At the same time,the adaptive reference frequency is extended to the second order for adapting the signals with fast-varying frequencies.Then the adaptive FSST2 algorithm with higher timefrequency aggregation is proposed.In view of insufficient knowledge in practical applications,the thesis proposes to use Rényi entropy for time-varying parameters estimation.By applying the proposed algorithms to analyze the synthetic and real signals respectively,and comparing with traditional methods,the proposed algorithms are proved to be effective and reliable.Secondly,in view of the insufficient estimation accuracy and poor robustness of commonly used IF estimation algorithms,this thesis proposes to combine the adaptive time-frequency SST and the Viterbi algorithm to estimate IF,which makes the proposed algorithm have the characteristics of high resolution and dynamic programming.Meanwhile,the multicomponent signals are classified according to whether there is crossover between the component IFs,and corresponding improvements are made to the proposed algorithm to further reduce the error of IF estimation.The experiments on different synthetic signals demonstrate the correctness and robustness of the proposed algorithms.Finally,this thesis applies the proposed algorithms to the parameter estimation of frequency hopping signals to verify its practicability.Aiming at the defect of large estimation error on STFT,the thesis uses the adaptive FSST and Viterbi algorithm to estimate the IFs of the frequency hopping signals.And the hop duration is estimated based on the the energy envelope spectrum of time-frequency analysis result to further estimate the hop time.At the same time,according to the advantages of chirp-z transform(CZT),the thesis proposes to refine the frequency spectrum by CZT to estimate hop frequencies precisely.Theoretical analysis and experimental results consistently verify that the proposed algorithms have better robustness.