Noise Source Identification Based On Compressed Sensing Method Study

Beamforming is an array-based sound source localization measurement technology that is widely used in the field of noise source identification.However,the traditional beamforming methods represented by delay and sum beamforming have low resolution and are only suitable for identifying high frequency sound sources.Because of the limits from Nyquist-Shannon sampling theory,we need to support a larger microphone array in order to get higher spatial resolution of sound source identification.To this end,this thesis introduces compressed sensing into the field of sound source identification.Compressed sensing is an "economic" sampling method that can sample at a lower Nyquist rate.If the signal is sparse in the transform domain,we only need a small amount of sampled data to achieve sound source localization.Aiming at the problem that the orthogonal matching pursuit algorithm in compressed sensing method is susceptible to coherence between atoms in sound source identification,an improved method is proposed to improve the recognition accuracy of sound source and the recognition ability of low frequency sound source.The specific research work is as follows:(1)This chapter reviews the development of sound source localization technology and compressed sensing principle,and introduces compressed sensing technology from three aspects: sparse signal representation,measurement matrix design and sparse reconstruction algorithm.(2)The chapter two introduces the L1 norm minimization algorithm into field of sound source localization.Aiming at the selection of regularization parameters in the L1 norm minimization algorithm,an auxiliary surface method is proposed to screen out the optimal regularization parameters.The numerical simulation results show that the sound source identification method based on L1 norm minimization algorithm has higher positioning resolution and recognition ability of medium and low frequency sound sources,and can accurately locate the sound source with less sampling data.(3)The chapter three introduces the orthogonal matching pursuit algorithm into the field of sound source localization.The orthogonal matching pursuit algorithm is widely used due to its fast convergence speed and high computational efficiency.This chapter uses orthogonal matching pursuit algorithm for sound source identification to obtain high sound source imaging accuracy.And the accurate positioning of the sound sourcecan be achieved with less sampled data.However,the orthogonal matching pursuit algorithm has the defect of weak recognition ability for multiple adjacent sound sources and low-frequency sound sources.(4)In order to solve problem that the orthogonal matching pursuit algorithm is susceptible to inter-atomic coherence,an improved orthogonal matching pursuit algorithm is proposed.The improved orthogonal matching pursuit algorithm adopts new atomic screening criteria to improve the reconstruction performance of the algorithm in a coherent environment.Through these improvements,the new algorithm achieves accurate recognition of low-frequency sound sources and achieves higher spatial resolution.Finally,the effectiveness of the sound source localization method based on improved orthogonal matching pursuit algorithm is verified by numerical simulation and experiment.