Continuous Compressive Sensing DOA Estimation In Spherical Harmonic Domain

The Direction of Arrival(DOA)estimation is one of the main aspects of array signal processing,and plays an important role in the fields of radar,communication and speech processing.Because of its symmetric character and full coverage of the whole three-dimensional angular space of the spherical array,the wide attention of the researchers has been obtained in recent years,and the research of the method of DOA estimation of the spherical harmonic domain has been deeply studied.The method for DOA estimation in spherical harmonic domain based on the discrete compressed sensing(CS)has good performance under the complex environment such as small number of snapshots,multipath propagation,but the methods have the drawbacks of grid mismatch error,large amount of calculation of multi-dimensional parameter optimization,and thus the practical application will be limited.However,the existing continuous compressed sensing DOA estimation methods based on atomic norm optimization rely on the Vandermonde structure of the array signal model,so they are only suitable for linear arrays.In view of the above problems,we extend the continuous compressed sensing DOA estimation method to spherical harmonic domain,and discuss the application of microphone array sound source location in room reverberation scenario.The main research contents are as follows:1.A L1-SVD-based DOA estimation method for the continuous compressed sensing spherical array is proposed in the paper.This paper establishs the relationship between the steering vector in spherical harmonic domain and the Vandermonde structural matrix,and extends the traditional atomic norm to spherical harmonic domain.On this basis,in the case of large snapshot,the L1-SVD method is used to reduce the dimension,and the two-dimensional DOA estimation of spherical array is realized.Compared with the traditional discrete compressed sensing method,The proposed method does not require meshing and high-dimensional multi-parameterization.It also has good direction finding performance and the ability to process coherent signals.2.It is difficult to select regularization parameters for DOA estimation method based on L1-SVD.In order to solve these problems,a continuous compressed sensing spherical array DOA estimation method based on covariance matrix is proposed in this paper.The spherical harmonic atomic decomposition and atomic norm optimization of covariance matrix are defined by taking the covariance matrix model in spherical harmonic domain as a multiple measurement vector(MMV).A Semidefinite Programing(SDP)method is given to solve the problem.The two-dimensional DOA estimation is realized by using the SH-ESPRIT method,which improves the robustness of the algorithm.3.One of the main challenges in locating multi-sources in the real world is the need to overcome the multi-path distortion caused by spatial reverberation.Based on the research of continuous compressed sensing DOA estimation method in spherical harmonic domain,this paper presents a spherical harmonic continuous compressed sensing method for locating sound source based on Direct-Path Dominance Test(DPD),which is applied to sound source location in room reverberation environment.This paper introduces frequency smoothing(FS)to reduce the influence of coherent signal source,and uses DPD Test to select TF-bins,which is dominant in direct path,to solve the problem of system underdetermination in reverberation environment,and then realizes sound source localization based on continuous compressed sensing DOA estimation method in spherical harmonic domain.In order to verify the effectiveness of this method,the computer simulation of typical room reverberation environment is carried out,and the performance of the method is compared with that of the traditional method.On this basis,the performance of the algorithm is verified for single or multiple static speakers scenes in the real room reverberation environment by using the experimental data of the LOCATA competition.The results show that the proposed method has better positioning accuracy and robustness than the existing methods in indoor reverberation environment.