Variational Bayes Based Spectral Estimation Algorithm And Application Software Implementation

Spectrum technology is widely used in the field of modern radar sonar and wireless communication.How to estimate the signal spectrum accurately and efficiently has become one of the most important research topics in the field of military and civil communication investigation.The purpose of this paper is to study the high resolution linear spectrum estimation and its application in passive sensing in underwater acoustic environment.First of all,aiming at the problem that the estimation accuracy is not high based on the classical power spectral density estimation,the iterative adaptive approach(i.e.IAA),the sparse learning via iterative minimization(i.e.SLIM)and the variational Bayesian interface of line spectra(i.e.VALSE)are studied.Simulation analysis and comparative analysis are carried out respectively.The simulation results show that,compared with the classical power spectral density estimation,IAA,SLIM and VALSE spectrum estimation methods make full use of the existing observation data,and improve the spectral estimation effect.IAA,SLIM and VALSE have higher accuracy,and the VALSE algorithm has the best performance of spectrum estimation among the three algorithms.Secondly,based on the experimental and data processing requirements,a cross-platform interface library wxWidgets is used to develop an application software based on Windows platform,which includes simulation,test data preview,time-spectrum analysis and playback.In the simulation mode,the signal length,frequency,sampling rate and other parameters of the single frequency signal and the linear frequency modulation signal can be set to generate the simulation signal,and the spectrum estimation simulation can be carried out under the background of white Gaussian noise.The spectrum estimation algorithm uses the VALSE algorithm which has the best performance in the spectrum estimation algorithm.In the test mode,the software reads the test data by reading the waveform file,analyzes the time spectrum of the test data,displays the waveform chart to the time frequency chart,and realizes the switching between the waveform chart and time-frequency chart by switching menu.For the simulation signal data and experimental data,the software can achieve data waveform preview,time-frequency analysis,play,stop,playback and jump to the beginning operation.Finally,each spectrum estimation method is used to deal with the experimental data,and the classical power spectral density estimation method is used to compare and analyze the experimental data.The experimental data analysis shows that the variational Bayesian interface of line spectra algorithm has high spectral estimation accuracy and the developed software can meet the design requirements.