Algorithm Improvements Of Sound Source Identification And Sound Field Separation Based On Equivalent Source Method

Near-field acoustic holography(NAH)based on microphone array measurement is an advanced sound field visualization technology,which is also widely used in automobiles and other fields.In a variety of NAH algorithms,the NAH based on the equivalent source method(ESM)has been widely studied and applied due to its ability to adapt to arbitrary-shaped sound sources and random arrays,as well as high reconstruction accuracy and computational efficiency.The most important issue of the ESM-based NAH is to solve the equivalent source amplitude.The reconstruction accuracy and computational efficiency based on different algorithm will vary.In this paper,three classical algorithms including Tikhonov regularization,CVX toolbox and wideband acoustic holography(WBH)are compared and studied,aiming at proposing an algorithm with higher reconstruction accuracy and computational efficiency.After that,double-layer sound field separation technique based on ESM is also improved.In order to improve the reconstruction accuracy of WBH,the threshold function and the initial value of the iteration are modified.Then,the idea of high-order matrix function beamforming is introduced to further improve the dynamic range of the sound source recognition results and improve the positioning accuracy of the algorithm.Further,aiming at the limitation of WBH inaccuracy of the recognition results of coherent sound sources in low and medium frequency,the monotonic two-step iterative contraction threshold algorithm(MTw IST)is used to solve the equivalent source amplitude vector in the framework of compressed sensing.The numerical simulation was carried out on single source,equal-amplitude/non-equal-range coherent sound source.The reconstruction accuracy and computational efficiency of the above four algorithms were compared.The simulation results show that MTw IST is close to the CVX toolbox in the available frequency range,and far better than the Tikhonov regularization and WBH;the computational efficiency is lower than the Tikhonov regularization,but it still has obvious advantage than the CVX toolbox.Finally,the correctness and feasibility of MTw IST were verified through experiments.The ESM-based NAH algorithm mentioned above requires that all sound sources be located on one side of the array.When there is an interference sound source on the back of the array,it cannot be accurately identified.It is necessary to use a sound field separation method to separate the sound source radiation field of the target sound source.In order to improve the separation accuracy of the conventional ESM-based pressure-pressure separation method,the sound source positions on both sides of the array are first determined using the double-sided sound pressure data and the iterative algorithm in WBH.Then a certain number of equivalent sources are arranged near the identified location,and condition number of the matrix is used as the goal for optimization with genetic algorithm.The coordinates of each equivalent source is optimized to achieve the purpose of improving the separation accuracy.Then,numerical simulations are used to prove the advantages of the improved method in separation accuracy.Finally,the correctness and effectiveness of the proposed method are verified by experiments.