Near Field Source Estimation Based On Acoustic Vector Sensor

Acoustic vector sensor(AVS)is a kind of sensor that can measure pressure and velocities at the same time.Compared with the traditional pressure sensor,AVS can get more information from the acoustic source signal.As the core component of sonar,AVS has been widely applied in the field of underwater acoustic engineering.Array signal processing is an important branch of signal processing,Scholars have also combined the knowledge of array signal processing to the acoustic vector sensor,which is used to locate the AVS parameters.Most of the existing estimation algorithms mainly focused on the situation in the far field,there is little research for the orientation estimation of near-field sound source.Near field acoustic source localization requires estimation of the source's direction of arrival and distance parameters,the problem of parameter estimation is more complex.Therefore,the far-field location method can not be directly applied to the localization of near-field source parameters.In this paper,the basis of quaternion theory and AVS ambiguity resolution algorithms are proposed for the far field receiving model.Near field acoustic source localization method based on compressed sensing,support vector regression and parallel factor method,the specific content as follows:Firstly,for the far field receiver model,when array interval vector more than half wavelength,the parameter estimation phase ambiguity problems emerged.Based on the theory of quaternion established the AVS array output signal model of quaternion,the quaternion AVS algorithm for solving phase ambiguity method has proposed.The quaternion AVS algorithm for solving phase ambiguity method has the advantages of smaller computational complexity and higher estimation precision than the long-vector for solving phase ambiguity method.The simulations of the uniform circular array are also conducted.The simulation results verify the effectiveness of the proposed algorithm.Then,for the near field receiver model,in view of the computational complexity problem of existing acoustic vector sensor near field higher-order ESPRIT algorithm,a new parameter estimation algorithm based on fourth-order cumulant and parallel factor analysis technology is proposed.The algorithm reduced the computational complexity.Basis on compressed sensing theory,using the thought of separating the azimuth parameter and range parameter of sources,three sparse restoration algorithms SOMP,L1-SVD and MFOCUSS are used.Their performance in near field acoustic source localization is analyzed and compared.A new algorithm based a cyclic three order moment combined with sparse recovery algorithm for DOA and range estimation of near-field cyclostationary sources is proposed in this paper.Basis on machine learning theory,a near field acoustic source localization algorithm based on MSVR is proposed.Compared with the two step MUSIC algorithm,the new algorithm can deal with coherent sources and resolution of small spaced angles at low signal-to-noise ratio(SNR).And the trained model can process of data quickly.The effectiveness of the proposed algorithm is verified by simulation.Finally,the paper a small software is made.The interface can be very convenient to modify the simulation parameters.These parameters mainly include the signal-to-noise ratio(SNR),the number of array element,the number of snapshot,angle and distance parameters of signal source,etc.And also can be convenient to show the simulation such as the root mean square error diagram,scatter diagram and probability of success diagram,the comparison graphics of various algorithms can also be showed.