Study On Acoustic Source Localization Based On Non-synchronous Measurement With Mobile Array

Acoustic source localization includes measuring acoustic data in sound field with sensor array and reconstructing acoustic source distribution image in sound field with back propagation algorithm.At present,acoustic source localization is widely used in industrial production,medical treatment,geological research and military fields.Receiving signal with microphone array is a key step in the process of acoustic source localization,and its sampling quality directly affects the subsequent accuracy of acoustic source localization.According to the sampling theorem,the upper limit of the received signal frequency of a microphone array is affected by the minimum distance two elements,and the lower limit is affected by the array size.Therefore,for the signal beyond the frequency limit,the quality of microphone sampling will decrease,which will reduce the accuracy of the localization of the acoustic source.However,in order to make the acoustic source localization algorithm suitable for the location of a larger frequency range,it is not practical to expand the size of a single microphone array without restriction,or to shrink the spacing of the array elements without considering the size of the elements.In order to solve this problem,the non-synchronous measurement acoustic source acquisition method is used in the thesis.The acoustic data is received at each preset measurement point with a mobile basic microphone array.By means of the matrix completion algorithm,the virtual union of the arrays at each measurement point and the virtual synchronization of signal reception are realized.Furthermore,the limitation of single microphone array size and array elements spacing on the frequency range of localizable acoustic source is solved.At the same time,through the research on the existing nonsynchronous measurement source localization algorithm,it is considered that,on the one hand,the acoustic source imaging resolution and localization accuracy of conventional beamforming acoustic source are affected by the microphone array.While in the expansion of virtual microphone array for non-synchronous measurement source localization,the location distribution of measurement points is an important element to determine the expanded array.On the other hand,the model and algorithm of cross spectral matrix complement affect the accuracy of the acoustic data combination and synchronization of each measurement point,and then affect the accuracy of the location of the sound source.In order to optimize the source localization algorithm of asynchronous measurement,these two aspects mentioned above are studied in this thesis.The details are as follows:(1)Aiming at the optimization of measurement point location,the influence of different measurement point distribution on acoustic source location is studied in this thesis by means of a group of simulation experiments.And peak side lobe level is found to be the main link and control factor between measurement point distribution and acoustic source imaging resolution.According to the optimization model of firefly algorithm in swarm optimization algorithm,a measurement point location optimization algorithm which takes the measurement point location coordinates as the optimization population and the peak sidelobe level as the fitness function is proposed in the thesis.The simulation results show that the proposed algorithm can suppress the sidelobe of sound source imaging,improve the accuracy of sound source location and expand the frequency range of sound source location.(2)Aiming at the optimization problem of cross spectral matrix completion model,the nuclear norm is used to approximate the low rank condition of cross spectral matrix in the existing matrix completion model of non-synchronous measurement acoustic source localization direction.The nuclear norm reduces the value of all eigenvalues of cross spectral matrix in the process of completion,which reduces the accuracy of cross spectral matrix completion.In order to better approximate the low rank condition,a cross spectral matrix completion model using truncated nuclear norm instead of kernel norm is proposed in this thesis.The model can selectively minimize the eigenvalues which do not correspond to the acoustic source signals,and then improve the accuracy of cross spectral matrix completion.(3)Aiming at the optimization problem of cross spectral matrix complement algorithm,new cross spectral matrix complement algorithms for non-synchronous measurement acoustic source localization are proposed in the thesis.The new algorithms which are based on the truncated kernel norm cross spectral matrix complement model implement the matrix complement by means of texture block matrix,matrix weighting technology and optimization algorithm in image processing field.The simulation results show that the proposed cross spectral matrix completion algorithm is effective in improving the upper limit of acoustic source location frequency and the accuracy of acoustic source location.