Research On Acoustic Source Localization Of Multi-node Information Fusion Based On WASN

Acoustic Source localization is of great significance in the field of signal processing,and has been widely used in many fields.With the development of microphone array and wireless sensor networks,the technology of wireless acoustic sensor network(WASN)has increasingly become a hot research as a new method of sound signal acquiring and processing.Based on the characteristics of WASN and under the application requirements,the acoustic source localization method of multi-node information fusion is studied in this thesis.The background significance,location advantages and research status of acoustic source localization methods based on WASN are summarized systematically.Then the theoretical basis of microphone array and sound source localization is introduced,in which the existing classical methods of source localization and the parameter estimation method of the near and far field source on the node are emphatically described.Thus,the limitation is pointed out that the existing methods only consider the far field condition,and the fusion center does not exist the fusion of near and far field estimation parameters.Aiming at the problem,the thesis focuses on the multi-node data fusion method for the fusion center containing far and near field source information and puts forward two kinds of solutions,a fusion method based on Kalman filter and another fusion method based on weighted K-means clustering.Finally,the validity and accuracy of the proposed algorithms are verified by simulation experiments,and the influence of the important parameters is discussed in detail.According to the shortage of existing cross positioning method,the algorithm of multi-node information fusion method that based Kalman filter for source localization is proposed in chapter three.Firstly,the improved weighted least squares method is used to the fusion of far field DOA preliminary and then the iterative Kalman filter method is utilized for noise elimination,further to realize the effective fusion of far and near field data.Through extending the target observation range on the basis of the grid-based method,the noise immunity is enhanced greatly and the positioning accuracy can be significantly improved by the combination of analytic method and iterative method.Besides,on account of the large amount of far field nodes with more error data and redundant information,another algorithm that a fusion method based on weighted K-means clustering is put forward in chapter four.The improved K-means clustering method is introduced for the data processing of far field,and by combining the density weighted method to achieve the far and near field data fusion.The method effectively eliminates the gross error data by the pretreatment of initial sample set and the post-processing of clustering result set,which makes full use of the advantages of K-means clustering.Moreover,it also does not depend on any system model and not subject to the application environment,which not only ensures the positioning performance but also expands the scope of application.The simulation results show that the proposed two algorithms are effective in the localization of multi-node information fusion,and have a good performance in the positioning accuracy,fault tolerance,stability and so on.