An Investigation On Two-dimensional Sparse Direction Finding Method With Total Variational Regularization And Its Experiments For Spatially Extended Sound Sources

Direction of Arrival(DOA)estimation has been widely used in radar,sonar,speech processing and wireless communication.With the emergence of practical requirements such as sound source type,acoustic environment complexity and application scene diversity,the requirements for DOA estimation accuracy,estimation dimension and resolution of direction finding system are getting higher and higher.Among them,the sparse reconstruction DOA estimation algorithm has become one of the hot research topics because it can better adapt to the modern direction finding system characterized by small snapshots,low signal-to-noise ratio and coherence while having ultra-high angle resolution.In practice,most of the sound sources are extended sound sources with obvious block sparsity and spatial continuous angle distribution characteristics,but the related DOA estimation results are less and basically limited to one dimension.In addition,the existing point source algorithm is also mismatched due to the strong correlation of the microphone measurement matrix and the stage linear characteristics of the observation signal due to the extended sound source.Therefore,in order to describe the spatial characteristics of the target more accurately,based on the principle of sparse reconstruction algorithms,this paper proposes a two-dimensional DOA estimation method based on total variation regularization(2DTV-CES).The main work is summarized as follows:(1)Based on the basic principle of DOA estimation of array signals,the principle of sparse reconstruction algorithm and the receiving model of related signals are studied.The principles,advantages and disadvantages of various classical subspace algorithms and regularization algorithms are analyzed.(2)For the spatial extended sound source,the mechanism of its correlation and block sparsity is studied and analyzed,and its expression form is defined and the extended sound source model is constructed.(3)Aiming at the performance degradation of traditional sparse DOA estimation methods in estimating extended sound sources,a two-dimensional sparse DOA estimation method based on total variation regularization is proposed.On the basis of constructing a two-dimensional array manifold matrix satisfying the corresponding constraints and its over-complete representation to realize the sparse representation of the sound source,a variational structure matrix is constructed.The continuous difference constraint of the coefficients of the total variational model promotes the formation of the piecewise constant contour of the solution.Combined with the sparse constraint term,the two-dimensional DOA estimation of the extended sound source is realized.(4)Numerical simulation of 2DTV-CES method.Taking a variety of extended sound sources as recognition objects,the numerical simulation results verify the spatial angle estimation ability and universality of the proposed method.Compared with other classical methods under different SNR and sparsity conditions,the numerical simulation results verify the robustness and superiority of the proposed method.(5)Experimental study of 2DTV-CES method.In the ideal semi-anechoic chamber environment and the complex external field environment,a variety of actual extended sound source information is collected.After the data is preprocessed,the DOA estimation results obtained by the proposed method are mutually confirmed with the numerical simulation part,which further verifies the practical applicability of the method.