A Robust High-Resolution Passive Positioning Method Based On Deconvolution For Near-Field Sound Source

The acoustic image measurement algorithm is widely used in the near-field noise source localization.The conventional acoustic image measurement algorithm has a large focus peak scale,which makes it difficult to locate the low frequency sound source.The work of this paper mainly focuses on the high resolution problem of acoustic image.The high resolution acoustic image measurement algorithm based on MVDR and deconvolution is studied respectively,and the deconvolution algorithm is accelerated by spatial resampling.Through simulation analysis and sea trial data,the practicality of the algorithm is verified.Firstly,an acoustic image measurement model for underwater noise source localization is established.The acoustic image measurement algorithm based on conventional focused beamforming is studied,and the factors affecting the focus peak scale of the acoustic image are analyzed.Then,the high-resolution acoustic image measurement algorithm based on MVDR and STMV beamforming is studied.The performance of the conventional acoustic image measurement algorithm and the high-resolution acoustic image measurement algorithm are compared by simulation experiments.Then the convolution model of acoustic image measurement is established,and the problems and difficulties of near-field deconvolution solving are studied.The shift-variant deconvolution acoustic image algorithms based on DAMAS and NNLS is studied because of the shift-variant point spread function in this convolution model.A predefined dictionary method is proposed to improve the Richardson-Lucy iterative algorithm,which makes it suitable for the near-field model.Aiming at the problem of large amount of computation of the shift-variant two-dimensional deconvolution algorithm,based on the prior knowledge acquired by conventional acoustic image measurement,a spatial resampling accelerated deconvolution algorithm is proposed,which effectively and reasonably reduces the spatial sampling points,thus reducing the amount of computation of the algorithm.The simulation results demonstrate the feasibility of three deconvolution algorithms and spatial resampling algorithms.Finally,the performance parameters of the acoustic image measurement algorithm are studied.The convergence of three deconvolution algorithms and the spatial resampling deconvolution algorithm are analyzed.The results show that the three deconvolution algorithms are convergent and the space resampling algorithm can accelerate the convergence of the deconvolution algorithm while reducing the amount of computation.The robustness of the deconvolution algorithm and the MVDR algorithm are analyzed.The results show that all the deconvolution algorithms are not sensitive to the array distortion.The Richardson-Lucy algorithm has the best comprehensive performance under the low SNR.Due to the array manifold vector mismatch,the MVDR algorithm has severe performance degradation under array distortion and low SNR.The sea trial data processing results are consistent with the theoretical analysis and simulation experiments.In summary,this paper studies a variety of high-resolution acoustic image measurement algorithms.The simulation and sea test results show that the deconvolution acoustic image measurement algorithm has high resolution and good robustness,and can effectively improve the positioning performance of low-frequency noise sources.