Research On Raypath Separation Technology Based On Compression Sensing

An important property of ocean acoustic tomography is the multipath propagation of acoustic signals in water.The first step of ocean acoustic tomography is the use of array processing technology to separate the raypaths and estimate the arrival time of different raypaths.The acoustic signal propagates along the multipath in the shallow sea,forming raypaths with different angles.To achieve DOA estimation means the separation of the raypaths is completed.However,the existing raypath separation algorithm still has problems such as low resolution and poor robustness to noise in complex environment.The recently proposed compressive sensing algorithms offer a new method for signal reconstruction and parameter estimation in the area of DOA estimation.In this thesis,two compressive sensing based algorithms are proposed in the background of the shallow water for effectively separating raypaths in the spatial domain.The main contributions of this thesis are illustrated as follows:(1)The basic principle of compressive sensing is elaborated.The classical narrowband DOA estimation method l1-SVD algorithm and l1-SRACV algorithm based on compressive sensing are analyzed theoretically,and simulation experiments were carried out The advantages and disadvantages of the existing methods are analyzed,which lays the theoretical foundation and research direction for extending the narrowband algorithm to the broadband field.(2)Aiming at the shortcomings of l1-SRACV algorithm,a wideband convex optimization algorithm based on array covariance matrix is proposed.The method decomposes the wideband signal into J narrowband components in the frequency domain,and then performs narrowband processing directly on each subband,which means,the atomic structure of the covariance matrix of the sampled signal is compared with each other,then the atomic structure is extracted partly.And the spatial spectrum of all subbands is averaged according to the orthogonality of the signal subspace and the noise subspace.Finally,the spatial spectrum of the broadband signal is estimated.The method of constructing the sparse matrix of the main diagonal elements of the covariance matrix obviously improves the shortcomings of the l1-SRACV algorithm.The simulation results show that the proposed method can effectively suppress the pseudo spectrum peaks and obtain more accurate DOA estimation.(3)Aiming at the problem that the existing sparse reconstruction DOA estimation algorithm can not effectively suppress the noise and in the case of coherent signal source,the covariance method can not accurately estimate the direction of the arrival wave,a wideband convex optimization algorithm based on fourth order cumulant is proposed.In the array direction estimation,the high-order cumulant can be used to suppress the spatial Gaussian noise.Due to the influence of multipath propagation,if there is a coherent signal in the received signal,the rank of the autocorrelation matrix is reduced.Sparse reconstruction algorithm is insensitive to the coherent signal.Based on the above two points,the sparse representation model is constructed by using the fourth-order cumulant matrix of array output data.By constructing the reduced fourth-order cumulant matrix,the redundant information of the elements in the traditional fourth-order cumulant matrix is eliminated,which greatly reduces the computational complexity.Compared with the broadband convex optimization algorithm based on covariance matrix,the algorithm not only applies to non-correlation signals,but also can directly deal with the relevant signal without any de-correlation operation.The simulation results show that the proposed method has obvious advantages in terms of computation and estimation accuracy.