The Method To Estimate Instantaneous Frequency Of Multicomponent Signals Based On Viterbi Algorithm

Instantaneous frequency can reflect the frequency change trend of the signal at a certain moment,which is an indispensable part of time-frequency analysis.It is of great significance for the analysis and research of non-stationary signals.With the development of research,it is found that the signals encountered in practical applications are often not single component signals,but more complex multi-component signals.In the time-frequency plane,these components are not necessarily independent,but may overlap.These conditions are likely to cause instantaneous frequency estimation accuracy reduction or detection errors and other problems.How to estimate the instantaneous frequency of multi-component signals,how to deal with overlapping signals on time-frequency,how to separate multi-component signals have been the hotpots of scholars.Based on Viterbi algorithm and its improved algorithm,this thesis studies the instantaneous frequency estimation of multi-component signals.The main work of this thesis includes:1.The concept and development of instantaneous frequency are described,and some classical instantaneous frequency estimation methods are described in detail.2.Several common time-frequency analysis methods,including definition,characteristic and deformation,are introduced in detail.The performance,advantages and disadvantages of each method are analyzed by simulation results.3.Viterbi algorithm(VA)and its improved algorithm(IVA)are described in detail.The reasons for the insufficient robustness of IVA at time-frequency overlap are studied from the perspective of definition,and simulation analysis is carried out.Combined with the idea of track correlation,a new improved algorithm based on Viterbi algorithm is proposed,which establishes a new penalty function by making multi-point and multi-step prediction of the current moment.Experimental results show that this method can not only estimate and separate the instantaneous frequency of component signals,but also overcome the shortcomings of previous algorithms in robustness at time-frequency overlap.4.The energy distribution of the signal is studied from the perspective of the three-dimensional time-frequency graph,and the reason of the instantaneous frequency jump at the time-frequency overlap point of the algorithm is analyzed.From the perspective of time-frequency energy,an improved method for this problem is proposed.In this method,the energy information of a signal component around the time-frequency overlap point is used to estimate its energy at that point.When the known instantaneous frequency trajectory is removed,the energy information of the remaining components is retained as much as possible.Experimental results show that this method can prevent the interruption of time-frequency energy and effectively avoid the jump of instantaneous frequency trajectory at the time-frequency overlap point.5.Comparative analysis of each algorithm,the simulation results show that the proposed algorithm is more robust than VA and IVA in the problem of instantaneous frequency estimation of multi-component signals.In addition,the proposed algorithm has higher accuracy and stronger noise resistance under the same noise environment.