| The most effective method of noise reduction is to control the noise from the noise source position,and the sound source identification algorithm is studied in this paper to achieve a modern living environment with low noise and an industrial production environment.Among them,the compressed sensing sound source identification algorithm has high accuracy in identifying the medium-high frequency and long-distance sound sources and is widely used in noise identification and location of wind power generation,automobiles,airplanes,trains,high-speed trains,and other equipment.It’s advantage of using the sparsity of sound source signals to project a large amount of data into low-dimensional space,thus greatly compressing data,has been widely applied and studied by domestic and foreign scholars.At present,the compressed sensing algorithm used in sound source recognition relies too much on the prior knowledge of sparsity while ensuring high recognition accuracy.Based on the advantages of the compressed sensing algorithm in the field of signal recognition,this thesis comprehensively considers the sparsity information and reconstruction accuracy of the compressed sensing sound source algorithm and proposes an improved algorithm for the compressed sensing sound source recognition algorithm.The specific analysis and research contents are as follows:The mathematical model of sound source identification is established,and three research directions of sparse signal representation,measurement matrix design,and compressed sensing of sparse signal reconstruction algorithm are introduced.The orthogonal matching pursuit compressed sensing reconstruction algorithm and piecewise weak matching pursuit compressed sensing reconstruction algorithm are mathematically deduced and theoretically analyzed.Aiming at the fact that the atomic screening rule process of the orthogonal matching pursuit algorithm is too simple and the iteration times of the algorithm depend on the accurate information of sparsity,a second weak selection compressed sensing sound source identification algorithm is proposed.The primary atoms are obtained by threshold parameters independent of independent of sparsity information.The current sound source estimate is calculated using the primary atoms,and the reliability of the primary atoms is detected by secondary selection according to the iterative rule.Finally,the final sound source estimate is calculated by the secondary selected atom set.The numerical simulation analysis shows that the algorithm has higher recognition accuracy and more accurate sound pressure amplitude under the condition of low dependence on sparsity.To solve the problem that the accuracy of secondary weak selection compressed sensing sound source identification algorithm still needs to be improved on the basis of improving the accuracy of sound source identification,a sparse independent regularization compressed sensing sound source identification algorithm is proposed based on the regularization orthogonal matching pursuit algorithm using the sparsity parameter as the upper limit.In this paper,after the atomic set is classified based on the initial stage,the index atomic set with the largest correlation in each iteration is selected as the candidate index atomic set by the selection criterion of the average maximum energy,and finally the final sound source estimate is calculated.The numerical simulation results show that this algorithm has the advantages of simple steps,small amount of calculation,high sound source imaging accuracy and quality,etc.without sparsity as a priori knowledge.Compared with the second weak selection compressed sensing sound source identification algorithm,it has improved the frequency range and identification accuracy.To verify the feasibility of the practical application of the TSWOMP algorithm and SIRP algorithm in sound source identification,and to build an experimental environment for sound source identification,an 18-channel microphone array is used to collect sound field information of single and double sound sources,and the signal is subjected to Fourier transform to obtain sound pressure amplitude and phase,and the traditional beamforming and compressed sensing algorithms are compared.The experimental results verify the feasibility and accuracy of the TSWOMP algorithm and SIRP algorithm.At the same time,by changing the frequency and number of sound sources,the applicability of the TSWOMP algorithm and SIRP algorithm is verified. |
