High Resolution DOA Estimation Of Wideband Signals In Fractional Fourier Domain

Underwater signal processing technology is an essential technical means to develop and maintain the ocean.As an important branch of underwater signal processing,DOA estimation has always been a hot research topic.In practical engineering,signals are mostly broadband signals,while traditional DOA estimation methods are mostly based on narrow-band signals.Therefore,it is of great practical significance to study such problems as high resolution DOA estimation algorithm for broadband signals and performance degradation under low SNR.As one of the important methods of high-resolution DOA estimation,MUSIC algorithm has the problem of large computation.To solve this problem,an improved MUSIC algorithm is proposed in this paper,which uses the function properties of the uniform linear matrix model to construct the mirror radiation source,thus reducing the computation of the spectral peak search part by 50%.Compared with the traditional MUSIC algorithm,with the decrease of the search step length and the improvement of the estimation accuracy,the calculation time of the improved MUSIC algorithm is approximately 50% of that of the traditional MUSIC algorithm,and the estimation accuracy is close to that of the traditional MUSIC algorithm.Aiming at the problem that the proposed DOA algorithm based on the narrow-band signal model is not applicable to the LFM signal model in this paper,the LFM signal is focused in the fractional-order Fourier domain by using the feature of the rotating time-frequency surface of the fractional-order Fourier transform,and then the DOA estimation algorithm of the narrow-band signal is used to estimate the arrival direction of the LFM signal.In the fractional Fourier domain,the improved MUSIC algorithm is used to further reduce the computation,so as to realize the fast DOA estimation of LFM signal.The simulation compares the FRFT-MUSIC algorithm,and the improved FRFT-MUSIC algorithm also reduces the computing amount by nearly 50% in the MUSIC spectrum peak search part,which can realize the fast high-resolution DOA estimation of LFM signal.Aiming at the problem of performance degradation of improved frft-music algorithm and frft-music algorithm under low signal-to-noise ratio,in order to realize robust high-resolution DOA estimation under low signal-to-noise ratio and further improve the estimation performance,this paper tries a DOA estimation algorithm based on RBF neural network.A model training method using the covariance matrix of LFM signal in FRF domain as input and the noise subspace matrix as output is proposed.In view of the poor performance of traditional frft-music algorithm under low SNR,a model training method combined with integrated learning idea is proposed,and the estimated success probability of traditional frft-music algorithm under low SNR is improved by simulation.Finally,through the processing of Zhoushan sea test data,the FRFT-CBF algorithm,the traditional FRFT-MUSIC algorithm,the improved FRFT-MUSIC algorithm and the RBF neural network algorithm are compared and analyzed to verify that the two methods proposed in this paper also have the advantages of small computation,high estimation accuracy and strong robustness in practical applications.