Research On Compressed Sensing Sound Source Identification Method Based On Orthogonal Matching Pursuit

The harmful effects of noise pollution caused by mechanical equipment have become increasingly apparent,underlining the importance of accurately identifying the sources of noise to develop effective noise reduction strategies.Compressed sensing sound source identification algorithms that employ microphone arrays have found widespread applications in the identification of noise sources in domains such as automobiles,aircraft,and mechanical equipment,owing to their ability to surpass the limitations imposed by the Nyquist sampling theorem and accurately locate noise sources.The Orthogonal Matching Pursuit(OMP)algorithm is widely used in compressed sensing greedy algorithms due to its simplicity,small computational requirements,and ease of implementation.However,the OMP algorithm’s accuracy diminishes when detecting low-frequency noise,and requires a priori knowledge of the sparse configuration of sound sources,significantly limiting its engineering practicality.To address these issues,an improved method has been proposed that seeks to alleviate the reliance on sparse sound sources,expand the effective range of identification,and improve identification accuracy.The specifics of the research are outlined below:By establishing a mathematical model for sound source identification,the principle and calculation process of the OMP algorithm were investigated.Numerical simulations of single and double sound sources were conducted,comparing the performance of Conventional Beamforming(CBF)and OMP algorithms under various factors such as frequency,grid spacing,and sound source location.The results showed that,when compared to CBF,the OMP algorithm significantly enhanced the accuracy and spatial resolution of sound source identification.However,the OMP algorithm had erroneous recognition under conditions such as low frequency,narrow grid spacing,and edge sound sources,resulting from the strong correlation between adjacent columns of the transfer matrix.Furthermore,OMP algorithm still requires knowing the sparsity of sound sources beforehand,which can complicate its practical application.To address the shortcomings and limitations of the OMP algorithm,an improved orthogonal matching pursuit sound source identification algorithm based on iterative shrinkage threshold prior solution is proposed.This method uses the iteratively shrinkage-thresholding algorithm to determine the rough solution of the sound source strength and guide the selection of new atoms,thereby replacing the traditional maximum inner product rule,and the problem of easily selecting wrong atoms is overcome.Numerical simulations were conducted on single and double sound sources,comparing the performance of the two algorithms under the same frequency,grid spacing,sound source position,and number of sound sources.The results show that the improved algorithm can significantly improve identification accuracy,expand the effective frequency recognition range,and extend the recognition area,demonstrating better multi-source positioning capability.When OMP algorithm is used to solve the deconvolution linear inverse problem,it is difficult to specify the number of sound sources in advance,so a deconvolution sound source identification algorithm based on piecewise thresholding orthogonal matching pursuit is proposed.By thresholding the inner product and the least squares solution,the algorithm effectively removes pseudo and sidelobe atoms from the atomic support set,achieving similar performance to the OMP-DAMAS algorithm while eliminating the need for prior knowledge.Numerical simulation studies the applicable frequency,signal-to-noise ratio,measurement distance and sound source distance of the PTOMP-DAMAS algorithm,and verifies that the PTOMP-DAMAS algorithm can accurately identify and locate the sound source when the sparsity of the sound source is unknown,which has better engineering practicability.Using single and double speakers to simulate the sound source for comparative experiments.In the experiment,HBK company’s 18-channel microphone array is used to measure the data,and then the measured data is processed according to the sound source identification algorithm.The results show that,compared with OMP algorithm,the improved OMP algorithm with iterative shrinkage threshold prior solution can expand the applicable frequency range and have a wider recognition area;PTOMP-DAMAS algorithm can solve the problem of depending on the sparsity of sound source.Through the analysis of experimental data,the correctness and effectiveness of the improved algorithm are verified.