Source Number Estimation Method Based On Ensemble Empirical Mode Decomposition Under Low SNR

Direction of arrival(DOA)estimation has become a very important research direction in array signal processing.It has been widely used in communication,navigation,sonar,radar and electronic countermeasures.In the field of DOA estimation,assuming that the number of sources estimated is different from the real number,it will have a very serious impact on the performance of high-resolution spatial spectrum estimation methods.Therefore,the accuracy of DOA estimation is a very important graduate project.In many practical natural scenes,the scene noise is no longer white noise.Therefore,the research of source number estimation algorithm in colored noise background has better practical significance.However,when the number of sources is close to the number of array elements and the signal-to-noise ratio is low,the performance of classical source estimation algorithms will deteriorate or even fail in the space colored noise scenario.Therefore,it is better to estimate the number of signal sources in the case of low signal-to-noise ratio.In this paper,we found this method of estimating the number of signal sources based on ensemble empirical mode decomposition.This method uses ensemble empirical mode decomposition to extract the characteristics of array signals with different number of signal sources,and then uses Hilbert transform to extract the instantaneous phase characteristics of array signals,The extracted instantaneous phase features are put into BP neural network for multi classification prediction.This paper introduces the basic signal model of uniform circular array and the data model of each element,and introduces two classical DOA estimation algorithms,including music algorithm based on matrix eigenspace decomposition and ESPRIT algorithm.The main factors affecting the performance of source estimation are also analyzed.The performance of the classical source number estimation is compared in two different noise backgrounds,one is Gaussian white noise,the other is colored noise.The robustness of the classical source number estimation algorithm is compared by simulation experiments under the changes of snapshot number,signal-to-noise ratio and source number.Empirical mode decomposition(EMD)is very sensitive to noise and low signal-to-noise ratio(SNR)makes mode aliasing more complex when extracting array signal features.A method of source number estimation based on ensemble EMD and BP neural network is proposed.In the proposed method,Gaussian white noise with zero mean value is added to the empirical mode decomposition(EMD),and the signal components of different time scales are automatically mapped to the corresponding IMF through the spectrum uniform distribution of white noise.The extracted instantaneous phase characteristics of array signal are put into BP neural network for training,and the classifier model which can estimate the number of sources is obtained.The construction of EEMD + HT algorithm model,and the results and analysis of computer simulation experiment and RF laboratory data experiment of the algorithm.When SNR is low and source is equal to M-1 at most,m-uca is used to estimate the number of sources.M uniform circular array(UCA)are used to receive the far-field narrow-band signals.The far-field narrow-band array signals are processed by EEMD to obtain a series of IMF components and a residual component.Then the first three components are transformed by Hilbert transform to obtain the instantaneous phase.Finally,the extracted phase is extracted to obtain a series of phase features,The instantaneous phase feature is put into BP neural network for training to get the algorithm model.Then,two groups of comparative experiments are used to verify the performance of the proposed algorithm.One group of comparative experiments uses the software simulation data,and the other uses the data collected by the RF laboratory.The performance and robustness of the proposed algorithm are verified by comparing with other five algorithms.