Application Research Of Principal Component Analysis In Direction Of Arrival Estimation Of Uniform Linear Array

With the rapid development of array signal processing technology,the requirement for signal direction finding accuracy in various fields is getting higher and higher.In order to achieve more accurate source estimation,the direction of arrival estimation has become a current research hotspot.The shortcomings of high-resolution algorithms such as MUSIC algorithm and ESPRIT algorithm are that the DOA estimation error at low signal-to-noise ratio is relatively large,and it cannot be directly used for DOA estimation of coherent sources.In recent years,with the rapid development of statistics and intelligent algorithms,more and more statistical intelligent algorithms are also used in DOA estimation,such as principal component analysis,neural network algorithm,support vector regression algorithm,etc.The advantage of this kind of algorithm is that it can construct the signal estimation model through modeling,and it can include the influence factors such as noise and array arrangement error in the model through training,and it does not need to search the spectral peak of the divided Angle space,and it can estimate the wave orientation more quickly.This paper discusses and analyzes the key techniques of this kind of algorithm.Firstly,this paper studies the application of PCA algorithm in the classical high-resolution algorithm(MUSIC,SOMP),and studies the array PCA-MUSIC algorithm and PCA-SOMP algorithm.Because these high-resolution algorithms often require a large number of snapshots when performing DOA estimation,and each sampled data contains noise information.Therefore,the PCA algorithm is used to reduce the dimension of the sampled signal to reconstruct the signal.The reconstructed signal still retains the information of the original signal,but reduces the noise information in the original signal.The improved array PCA-MUSIC algorithm and array PCA-SOMP algorithm have improved DOA estimation performance under low signal-to-noise ratio.Secondly,this paper studies the application of PCA algorithm in neural network algorithm,and is used for parameter estimation of near-field sources.The array PCA-BP and array PCA-RBF algorithms are studied,and the input data is reduced through PCA,thereby reducing the number of input neurons,making the structure of the neural network simple and reducing the training time of the network.It also improves the estimation performance of near-field source parameters under low signal-to-noise ratio.At the same time,array PCA-BP algorithm can also accurately estimate the near-field source parameters in the case of array errors.Finally,this paper studies the application of the PCA algorithm in the regression algorithm,and is used to estimate the parameters of near-field sources.The partial least squares regression algorithm is studied to estimate the near-field source parameters,and the NSF-PLSR algorithm is studied,which can extract the characteristic variables of the signal,remove redundant information and noise information,and estimate the near-field source parameters with high accuracy.In addition,PCA dimensionality reduction multi-output support vector regression algorithm is also used to estimate the near-field source parameters.The PCA-NSF-MSVR algorithm reduces the input signal dimension through PCA,reduces the calculation amount of the algorithm,and removes more noise information.Simulation experiments show the effectiveness of the proposed algorithm.The PCA-NSF-MSVR algorithm can also estimate coherent near-field sources and has good estimation performance.