Matched Field Localization Based On Sparse Space

Matched Field Processing(MFP)is one of the classical problems in underwater acoustics and the means to locate underwater targets.Conventional matched field processing suffers from low resolution and high side lobe level.The existing matched field processing with high resolution suffer form poor tolerance,and they also greatly depend on the number of snapshots and coherent sources can be resolved by them.Compared with the conventional matched field processing,matched field processing based on compressive sensing can overcome these defects.So matched field processing based on compressive sensing,which has been widely concerned by researchers in many countries,is one of the promising directions in underwater acoustic technology.An entirely new theory of information acquisition which broke through the limitation of traditional sampling theorem is proposed by Compressive Sensing theory.It collects the sample signal by random sampling,and then reconstructs the signal through the nonlinear reconstruction algorithm.The theory further improved and enriched the theory of signal sparse representation.It is a significant change of modern information theory.In this paper,combined with the research achievements of Compressive Sensing theory,the method of matched field processing based on space sparse reconstruction is studied.Firstly,based on the sparsity of actual signal in MFP search space,this paper carried out the source localization research which based on Compressive Sensing theory with the combination of the MFP model.As we all know,Compressive Sensing theory often uses sparse representation model based on l1 norm in order to get answer conveniently.But the problem is that this model estimation precision is not high.Therefore,in this paper we provide a new matched field processing method based on l0 norm to reduce the high computational complexity of the model reconstruction.In view of the problem above,we also use another fast reconstruction algorithm,namely the matched field processing method based on smooth l0 norm,to demonstrate the feasibility and validity of this method through the simulation experiment.Secondly,we improve smoothed l0 norm algorithm to reduce the time of sparse reconstruction.This paper combines with the improved smoothed l0 norm algorithm and proposes a matched field processing method which is based on improved smoothed l0 norm algorithm.Instead of a Gaussian function used in smoothness l0 norm algorithm,the improved method uses a combination function with more steep to approximate l0 norm and it combines with modified Newton and damping step.It avoids "effect of sawtooth" ofexistence and the problems of slow convergence speed in the smoothed l0 algorithm which takes the method of steepest descent and the gradient projection.At the same time,the improved method makes up for the various problems existing in the traditional matched field processing methods.Finally,the method of matched field processing based on Compressive Sensing is mainly on the base of the spatial sparsity and does not take their intrinsic relationship between the sparse characteristics into consideration.However,when the sound source in matched field positioning is observed,the sparsity of the target signals will have a certain link and show structural sparse characteristics.Consequently,another work of this paper is to build a sparse representation model of MFP based on structure sparse,and propose a MFP method based on sparse structure.The simulation experiments show that the reconstruction precision is high and it has spatial resolution as well as good localization performance in the situation of low SNR in this way.