Algorithm Study Of Sound Sources Identification Based On Sparsity Regularization

The sound source identification method based on the microphone array signal processing is one of the effective ways to solve the problem of fault diagnosis in noise control.Sparse promotion technology has been recently widely applied to acoustic source recognition because of its superiority of using the sparse structure of signals to reduce sampling frequency and improve signal reconstruction.Its resolution is much higher than that of the traditional identification methods,which has a broad application prospect.A series of research on the method of sparsity promotion is carried out in this paper.The main contents are as follows:To improve the performance of l₁-GIB and WBH at middle-low frequencies,a method based on improved Bregman iteration is proposed in this paper.This algorithm aims at minimize the l₁-norm of sources strength,and breaks down the optimization problem into a consequence of unconstrained sub-problems.The reweighting function is merged to punish each sub-problem's objective function,and the sub-problems are calculated via fast iterative shrinkage algorithm such that the calculated value of the intensity of the sound is closer to the theoretical value.To improve the stability of the algorithm,the contraction curve of the iterative shrinkage algorithm is relaxed.Meanwhile,in order to reduce the size of the transfer matrix,the interpolated wavelet decomposition is introduced to the pre-extraction of the computational grid.Numerical simulation and experimental data verify the effectiveness of the proposed method for the recognition of monopole and dual coherent sound source.The proposed method broaden the applicable frequency range.To further improve the computational efficiency,a fast linear Bregman iterative source recognition method based on non-monotone line search is proposed based on the Bregman iterative sparse promotion algorithm.In order to overcome the limitation of measurement aperture,a method of iterative extrapolation of the holographic aperture data is combined to the improved method.The convergence of the extrapolation process of the improved method is validated by numerical simulation.The influence of different frequency and signal to noise ratio on the recognition results is further simulated and experimentally studied.The results show that the extrapolation method based on linear acceleration Bregman iteration is convergent,and the recognition effect of low frequency sound source is further improved.Finally,to overcome the drawback of planar microphone array based recognition method,which could not identify the deep spatial array-source distance.We propose a hybrid source identification method based on sparse promotion and multiplicative filtering.First,aiming at a better array mode,a spatial cross axis array is designed by using genetic algorithm.Then,the space crosshair is regarded as a combination of three sub arrays,and the hybrid algorithm is adopted respectively,and then the output of the sub array is multiplied for filtering.The simulation and experimental results verify the effectiveness of the proposed method.The results show that the source recognition technology based on sparsity promotes the application of spatial sound source recognition,and it has application prospect.